diff options
Diffstat (limited to 'xlators/mgmt/glusterd/src')
32 files changed, 15935 insertions, 1219 deletions
diff --git a/xlators/mgmt/glusterd/src/Makefile.am b/xlators/mgmt/glusterd/src/Makefile.am index a6f49ae01..933c44019 100644 --- a/xlators/mgmt/glusterd/src/Makefile.am +++ b/xlators/mgmt/glusterd/src/Makefile.am @@ -11,7 +11,9 @@ glusterd_la_SOURCES = glusterd.c glusterd-handler.c glusterd-sm.c \ glusterd-volgen.c glusterd-rebalance.c glusterd-quota.c \ glusterd-geo-rep.c glusterd-replace-brick.c glusterd-log-ops.c \ glusterd-volume-ops.c glusterd-brick-ops.c glusterd-mountbroker.c \ - glusterd-syncop.c glusterd-hooks.c glusterd-volume-set.c + glusterd-syncop.c glusterd-hooks.c glusterd-volume-set.c \ + glusterd-locks.c glusterd-snapshot.c glusterd-mgmt-handler.c \ + glusterd-mgmt.c glusterd_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \ $(top_builddir)/rpc/xdr/src/libgfxdr.la \ @@ -21,7 +23,8 @@ glusterd_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \ noinst_HEADERS = glusterd.h glusterd-utils.h glusterd-op-sm.h \ glusterd-sm.h glusterd-store.h glusterd-mem-types.h \ glusterd-pmap.h glusterd-volgen.h glusterd-mountbroker.h \ - glusterd-syncop.h glusterd-hooks.h + glusterd-syncop.h glusterd-hooks.h glusterd-locks.h \ + glusterd-mgmt.h AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \ -I$(rpclibdir) -I$(CONTRIBDIR)/rbtree \ diff --git a/xlators/mgmt/glusterd/src/glusterd-brick-ops.c b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c index 7c8e55bef..596503c21 100644 --- a/xlators/mgmt/glusterd/src/glusterd-brick-ops.c +++ b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c @@ -922,6 +922,7 @@ _glusterd_restart_gsync_session (dict_t *this, char *key, char *slave_vol = NULL; char *slave_ip = NULL; char *conf_path = NULL; + char **errmsg = NULL; int ret = -1; glusterd_gsync_status_temp_t *param = NULL; gf_boolean_t is_running = _gf_false; @@ -957,10 +958,13 @@ _glusterd_restart_gsync_session (dict_t *this, char *key, ret = glusterd_get_slave_details_confpath (param->volinfo, param->rsp_dict, &slave_ip, &slave_vol, - &conf_path); + &conf_path, errmsg); if (ret) { - gf_log ("", GF_LOG_ERROR, - "Unable to fetch slave or confpath details."); + if (*errmsg) + gf_log ("", GF_LOG_ERROR, "%s", *errmsg); + else + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave or confpath details."); goto out; } @@ -1021,6 +1025,8 @@ glusterd_op_perform_add_bricks (glusterd_volinfo_t *volinfo, int32_t count, glusterd_brickinfo_t *brickinfo = NULL; glusterd_gsync_status_temp_t param = {0, }; gf_boolean_t restart_needed = 0; + char msg[1024] __attribute__((unused)) = {0, }; + int caps = 0; GF_ASSERT (volinfo); @@ -1101,12 +1107,31 @@ glusterd_op_perform_add_bricks (glusterd_volinfo_t *volinfo, int32_t count, if (count) brick = strtok_r (brick_list+1, " \n", &saveptr); +#ifdef HAVE_BD_XLATOR + if (brickinfo->vg[0]) + caps = CAPS_BD | CAPS_THIN | + CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT; +#endif while (i <= count) { ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, &brickinfo); if (ret) goto out; +#ifdef HAVE_BD_XLATOR + /* Check for VG/thin pool if its BD volume */ + if (brickinfo->vg[0]) { + ret = glusterd_is_valid_vg (brickinfo, 0, msg); + if (ret) { + gf_log (THIS->name, GF_LOG_CRITICAL, "%s", msg); + goto out; + } + /* if anyone of the brick does not have thin support, + disable it for entire volume */ + caps &= brickinfo->caps; + } else + caps = 0; +#endif if (uuid_is_null (brickinfo->uuid)) { ret = glusterd_resolve_brick (brickinfo); @@ -1143,7 +1168,7 @@ glusterd_op_perform_add_bricks (glusterd_volinfo_t *volinfo, int32_t count, dict_foreach (volinfo->gsync_slaves, _glusterd_restart_gsync_session, ¶m); } - + volinfo->caps = caps; out: GF_FREE (free_ptr1); GF_FREE (free_ptr2); @@ -1240,15 +1265,6 @@ glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr) goto out; } - if (volinfo->backend == GD_VOL_BK_BD) { - snprintf (msg, sizeof (msg), "Add brick is not supported for " - "Block backend volume %s.", volname); - gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - ret = -1; - goto out; - } - ret = glusterd_validate_volume_id (dict, volinfo); if (ret) goto out; @@ -1326,6 +1342,18 @@ glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr) } if (!uuid_compare (brickinfo->uuid, MY_UUID)) { +#ifdef HAVE_BD_XLATOR + if (brickinfo->vg[0]) { + ret = glusterd_is_valid_vg (brickinfo, 1, msg); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "%s", + msg); + *op_errstr = gf_strdup (msg); + goto out; + } + } +#endif + ret = glusterd_validate_and_create_brickpath (brickinfo, volinfo->volume_id, op_errstr, is_force); @@ -1360,7 +1388,6 @@ glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr) glusterd_volinfo_t *volinfo = NULL; char *errstr = NULL; int32_t brick_count = 0; - int32_t replica_cnt = 0; char msg[2048] = {0,}; int32_t flag = 0; gf1_op_commands cmd = GF_OP_CMD_NONE; @@ -1417,10 +1444,10 @@ glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr) case GF_OP_CMD_START: { if ((volinfo->type == GF_CLUSTER_TYPE_REPLICATE) && - !dict_get_int32 (dict, "replica-count", &replica_cnt)) { - snprintf (msg, sizeof(msg), "Rebalancing not needed " - "when reducing replica count. Try without " - "the 'start' option"); + dict_get (dict, "replica-count")) { + snprintf (msg, sizeof(msg), "Migration of data is not " + "needed when reducing replica count. Use the" + " 'force' option"); errstr = gf_strdup (msg); gf_log (this->name, GF_LOG_ERROR, "%s", errstr); goto out; @@ -1453,7 +1480,7 @@ glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr) goto out; } - if (is_origin_glusterd ()) { + if (is_origin_glusterd (dict)) { ret = glusterd_generate_and_set_task_id (dict, GF_REMOVE_BRICK_TID_KEY); if (ret) { @@ -1640,15 +1667,6 @@ glusterd_op_add_brick (dict_t *dict, char **op_errstr) goto out; } - /* Need to reset the defrag/rebalance status accordingly */ - switch (volinfo->rebal.defrag_status) { - case GF_DEFRAG_STATUS_FAILED: - case GF_DEFRAG_STATUS_COMPLETE: - volinfo->rebal.defrag_status = 0; - default: - break; - } - ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); if (ret) goto out; @@ -1680,6 +1698,8 @@ glusterd_op_remove_brick (dict_t *dict, char **op_errstr) glusterd_brickinfo_t *tmp = NULL; char *task_id_str = NULL; xlator_t *this = NULL; + dict_t *bricks_dict = NULL; + char *brick_tmpstr = NULL; this = THIS; GF_ASSERT (this); @@ -1707,7 +1727,7 @@ glusterd_op_remove_brick (dict_t *dict, char **op_errstr) /* Set task-id, if available, in ctx dict for operations other than * start */ - if (is_origin_glusterd () && (cmd != GF_OP_CMD_START)) { + if (is_origin_glusterd (dict) && (cmd != GF_OP_CMD_START)) { if (!uuid_is_null (volinfo->rebal.rebalance_id)) { ret = glusterd_copy_uuid_to_dict (volinfo->rebal.rebalance_id, dict, @@ -1720,10 +1740,13 @@ glusterd_op_remove_brick (dict_t *dict, char **op_errstr) } } - /* Clear task-id & rebal.op on commmitting/stopping remove-brick */ + /* Clear task-id, rebal.op and stored bricks on commmitting/stopping + * remove-brick */ if ((cmd != GF_OP_CMD_START) || (cmd != GF_OP_CMD_STATUS)) { uuid_clear (volinfo->rebal.rebalance_id); volinfo->rebal.op = GD_OP_NONE; + dict_unref (volinfo->rebal.dict); + volinfo->rebal.dict = NULL; } ret = -1; @@ -1808,7 +1831,20 @@ glusterd_op_remove_brick (dict_t *dict, char **op_errstr) goto out; } - + /* Save the list of bricks for later usage. Right now this is required + * for displaying the task parameters with task status in volume status. + */ + bricks_dict = dict_new (); + if (!bricks_dict) { + ret = -1; + goto out; + } + ret = dict_set_int32 (bricks_dict, "count", count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to save remove-brick count"); + goto out; + } while ( i <= count) { snprintf (key, 256, "brick%d", i); ret = dict_get_str (dict, key, &brick); @@ -1818,6 +1854,21 @@ glusterd_op_remove_brick (dict_t *dict, char **op_errstr) goto out; } + brick_tmpstr = gf_strdup (brick); + if (!brick_tmpstr) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, + "Failed to duplicate brick name"); + goto out; + } + ret = dict_set_dynstr (bricks_dict, key, brick_tmpstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add brick to dict"); + goto out; + } + brick_tmpstr = NULL; + ret = glusterd_op_perform_remove_brick (volinfo, brick, force, &need_rebalance); if (ret) @@ -1831,6 +1882,7 @@ glusterd_op_remove_brick (dict_t *dict, char **op_errstr) volinfo->replica_count, replica_count, volinfo->volname); volinfo->replica_count = replica_count; + volinfo->sub_count = replica_count; volinfo->dist_leaf_count = glusterd_get_dist_leaf_count (volinfo); volinfo->subvol_count = (volinfo->brick_count / volinfo->dist_leaf_count); @@ -1847,6 +1899,8 @@ glusterd_op_remove_brick (dict_t *dict, char **op_errstr) } } } + volinfo->rebal.dict = bricks_dict; + bricks_dict = NULL; ret = glusterd_create_volfiles_and_notify_services (volinfo); if (ret) { @@ -1891,5 +1945,9 @@ out: if (ret && err_str[0] && op_errstr) *op_errstr = gf_strdup (err_str); + GF_FREE (brick_tmpstr); + if (bricks_dict) + dict_unref (bricks_dict); + return ret; } diff --git a/xlators/mgmt/glusterd/src/glusterd-geo-rep.c b/xlators/mgmt/glusterd/src/glusterd-geo-rep.c index 36a70e6fd..5786694bd 100644 --- a/xlators/mgmt/glusterd/src/glusterd-geo-rep.c +++ b/xlators/mgmt/glusterd/src/glusterd-geo-rep.c @@ -33,11 +33,32 @@ glusterd_get_statefile_name (glusterd_volinfo_t *volinfo, char *slave, char *conf_path, char **statefile); static int -glusterd_get_slave_info (dict_t *dict, char **slave_ip, char **slave_vol); +glusterd_get_slave_info (char *slave, char **slave_ip, + char **slave_vol, char **op_errstr); static int glusterd_gsync_read_frm_status (char *path, char *buf, size_t blen); +struct gsync_config_opt_vals_ gsync_confopt_vals[] = { + {.op_name = "change_detector", + .no_of_pos_vals = 2, + .case_sensitive = _gf_true, + .values = {"xsync", "changelog"}, + }, + {.op_name = "special_sync_mode", + .no_of_pos_vals = 2, + .case_sensitive = _gf_true, + .values = {"partial", "recover"} + }, + {.op_name = "log-level", + .no_of_pos_vals = 5, + .case_sensitive = _gf_false, + .values = {"critical", "error", "warning", "info", "debug"} + }, + {.op_name = NULL, + }, +}; + static char *gsync_reserved_opts[] = { "gluster-command-dir", "pid-file", @@ -781,15 +802,19 @@ out: static int gsync_verify_config_options (dict_t *dict, char **op_errstr, char *volname) { - char **resopt = NULL; - int i = 0; - int ret = -1; - char *subop = NULL; - char *slave = NULL; - char *op_name = NULL; - char *op_value = NULL; - char *t = NULL; - gf_boolean_t banned = _gf_true; + char **resopt = NULL; + int i = 0; + int ret = -1; + char *subop = NULL; + char *slave = NULL; + char *op_name = NULL; + char *op_value = NULL; + char *t = NULL; + char errmsg[PATH_MAX] = ""; + gf_boolean_t banned = _gf_true; + gf_boolean_t op_match = _gf_true; + gf_boolean_t val_match = _gf_true; + struct gsync_config_opt_vals_ *conf_vals = NULL; if (dict_get_str (dict, "subop", &subop) != 0) { gf_log ("", GF_LOG_WARNING, "missing subop"); @@ -863,6 +888,42 @@ gsync_verify_config_options (dict_t *dict, char **op_errstr, char *volname) } } + /* Check options in gsync_confopt_vals for invalid values */ + for (conf_vals = gsync_confopt_vals; conf_vals->op_name; conf_vals++) { + op_match = _gf_true; + for (i = 0; conf_vals->op_name[i] && op_name[i]; i++) { + if (conf_vals->op_name[i] == op_name[i] || + (conf_vals->op_name[i] == '_' && op_name[i] == '-')) + continue; + op_match = _gf_false; + } + + if (op_match) { + val_match = _gf_false; + for (i = 0; i < conf_vals->no_of_pos_vals; i++) { + if(conf_vals->case_sensitive){ + if (!strcmp (conf_vals->values[i], op_value)) + val_match = _gf_true; + } else { + if (!strcasecmp (conf_vals->values[i], op_value)) + val_match = _gf_true; + } + } + + if (!val_match) { + ret = snprintf (errmsg, sizeof(errmsg) - 1, + "Invalid values (%s) for" + " option %s", op_value, + op_name); + errmsg[ret] = '\0'; + + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + *op_errstr = gf_strdup (errmsg); + return -1; + } + } + } + return 0; } @@ -877,9 +938,9 @@ _get_status_mst_slv (dict_t *this, char *key, data_t *value, void *data) glusterd_gsync_status_temp_t *param = NULL; char *slave = NULL; char *slave_buf = NULL; + char *slave_ip = NULL; char *slave_vol = NULL; - char *tmp = NULL; - char *save_ptr = NULL; + char *errmsg = NULL; char conf_path[PATH_MAX] = ""; int ret = -1; glusterd_conf_t *priv = NULL; @@ -897,34 +958,26 @@ _get_status_mst_slv (dict_t *this, char *key, data_t *value, void *data) } slave = strchr(value->data, ':'); - if (slave) { - slave ++; - slave_buf = gf_strdup (slave); - if (!slave_buf) { - gf_log ("", GF_LOG_ERROR, - "Failed to gf_strdup"); - ret = -1; - goto out; - } - } else + if (!slave) return 0; + slave++; - tmp = strtok_r (slave_buf, ":", &save_ptr); - while (tmp) { - slave_vol = tmp; - tmp = strtok_r (NULL, ":", &save_ptr); - } - - if (!slave_vol) { - gf_log ("", GF_LOG_ERROR, "Invalid slave name"); + ret = glusterd_get_slave_info (slave, &slave_ip, &slave_vol, &errmsg); + if (ret) { + if (errmsg) + gf_log ("", GF_LOG_ERROR, "Unable to fetch " + "slave details. Error: %s", errmsg); + else + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave details."); ret = -1; goto out; } ret = snprintf (conf_path, sizeof(conf_path) - 1, - "%s/"GEOREP"/%s-%s/gsyncd.conf", + "%s/"GEOREP"/%s_%s_%s/gsyncd.conf", priv->workdir, param->volinfo->volname, - slave_vol); + slave_ip, slave_vol); conf_path[ret] = '\0'; ret = glusterd_get_gsync_status_mst_slv(param->volinfo, @@ -1337,11 +1390,11 @@ glusterd_verify_gsync_status_opts (dict_t *dict, char **op_errstr) } ret = glusterd_get_slave_details_confpath (volinfo, dict, &slave_ip, - &slave_vol, &conf_path); + &slave_vol, &conf_path, + op_errstr); if (ret) { - snprintf (errmsg, sizeof (errmsg), - "Unable to fetch slave or confpath details."); - gf_log ("", GF_LOG_ERROR, "%s", errmsg); + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave or confpath details."); ret = -1; goto out; } @@ -1349,7 +1402,6 @@ glusterd_verify_gsync_status_opts (dict_t *dict, char **op_errstr) out: gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; - } @@ -1403,6 +1455,22 @@ glusterd_op_stage_sys_exec (dict_t *dict, char **op_errstr) char command_path[PATH_MAX] = ""; struct stat st = {0,}; int ret = -1; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + GF_ASSERT (conf); + + if (conf->op_version < 2) { + gf_log ("", GF_LOG_ERROR, "Op Version not supported."); + snprintf (errmsg, sizeof(errmsg), "One or more nodes do not" + " support the required op version."); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } ret = dict_get_str (dict, "command", &command); if (ret) { @@ -1461,6 +1529,15 @@ glusterd_op_stage_copy_file (dict_t *dict, char **op_errstr) goto out; } + if (priv->op_version < 2) { + gf_log ("", GF_LOG_ERROR, "Op Version not supported."); + snprintf (errmsg, sizeof(errmsg), "One or more nodes do not" + " support the required op version."); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + ret = dict_get_str (dict, "host-uuid", &host_uuid); if (ret < 0) { gf_log ("", GF_LOG_ERROR, "Unable to fetch" @@ -1577,7 +1654,7 @@ glusterd_get_statefile_name (glusterd_volinfo_t *volinfo, char *slave, } static int -glusterd_create_status_file (char *master, char *slave, +glusterd_create_status_file (char *master, char *slave, char *slave_ip, char *slave_vol, char *status) { int ret = -1; @@ -1600,8 +1677,8 @@ glusterd_create_status_file (char *master, char *slave, runinit (&runner); runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "--create", status, "-c", NULL); - runner_argprintf (&runner, "%s/"GEOREP"/%s-%s/gsyncd.conf", - priv->workdir, master, slave_vol); + runner_argprintf (&runner, "%s/"GEOREP"/%s_%s_%s/gsyncd.conf", + priv->workdir, master, slave_ip, slave_vol); runner_argprintf (&runner, ":%s", master); runner_add_args (&runner, slave, NULL); synclock_unlock (&priv->big_lock); @@ -1621,13 +1698,15 @@ out: static int glusterd_verify_slave (char *volname, char *slave_ip, char *slave, - char **op_errstr) + char **op_errstr, gf_boolean_t *is_force_blocker) { - int32_t ret = -1; - runner_t runner = {0,}; - glusterd_conf_t *priv = NULL; - char log_file_path[PATH_MAX] = ""; - char buf[PATH_MAX] = ""; + int32_t ret = -1; + runner_t runner = {0,}; + char log_file_path[PATH_MAX] = ""; + char buf[PATH_MAX] = ""; + char *tmp = NULL; + char *save_ptr = NULL; + glusterd_conf_t *priv = NULL; GF_ASSERT (volname); GF_ASSERT (slave_ip); @@ -1662,7 +1741,26 @@ glusterd_verify_slave (char *volname, char *slave_ip, char *slave, log_file_path); goto out; } - *op_errstr = gf_strdup (buf); + + /* Tokenize the error message from gverify.sh to figure out + * if the error is a force blocker or not. */ + tmp = strtok_r (buf, "|", &save_ptr); + if (!strcmp (tmp, "FORCE_BLOCKER")) + *is_force_blocker = 1; + else { + /* No FORCE_BLOCKER flag present so all that is + * present is the error message. */ + *is_force_blocker = 0; + if (tmp) + *op_errstr = gf_strdup (tmp); + ret = -1; + goto out; + } + + /* Copy rest of the error message to op_errstr */ + tmp = strtok_r (NULL, "|", &save_ptr); + if (tmp) + *op_errstr = gf_strdup (tmp); ret = -1; goto out; } @@ -1674,6 +1772,72 @@ out: } int +glusterd_mountbroker_check (char **slave_ip, char **op_errstr) +{ + int ret = -1; + char *tmp = NULL; + char *save_ptr = NULL; + char *username = NULL; + char *host = NULL; + char errmsg[PATH_MAX] = ""; + + GF_ASSERT (slave_ip); + GF_ASSERT (*slave_ip); + + /* Checking if hostname has user specified */ + host = strstr (*slave_ip, "@"); + if (!host) { + gf_log ("", GF_LOG_DEBUG, "No username provided."); + ret = 0; + goto out; + } else { + /* Moving the host past the '@' and checking if the + * actual hostname also has '@' */ + host++; + if (strstr (host, "@")) { + gf_log ("", GF_LOG_DEBUG, "host = %s", host); + ret = snprintf (errmsg, sizeof(errmsg) - 1, + "Invalid Hostname (%s).", host); + errmsg[ret] = '\0'; + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + if (op_errstr) + *op_errstr = gf_strdup (errmsg); + goto out; + } + + /* Fetching the username and hostname + * and checking if the username is non-root */ + username = strtok_r (*slave_ip, "@", &save_ptr); + tmp = strtok_r (NULL, "@", &save_ptr); + if (strcmp (username, "root")) { + ret = snprintf (errmsg, sizeof(errmsg) - 1, + "Non-root username (%s@%s) not allowed.", + username, tmp); + errmsg[ret] = '\0'; + if (op_errstr) + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, + "Non-Root username not allowed."); + ret = -1; + goto out; + } + + *slave_ip = gf_strdup (tmp); + if (!*slave_ip) { + gf_log ("", GF_LOG_ERROR, "Out of memory"); + ret = -1; + goto out; + } + } + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr) { char *down_peerstr = NULL; @@ -1686,10 +1850,12 @@ glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr) char *conf_path = NULL; char errmsg[PATH_MAX] = ""; char common_pem_file[PATH_MAX] = ""; + char hook_script[PATH_MAX] = ""; char uuid_str [64] = ""; int ret = -1; int is_pem_push = -1; gf_boolean_t is_force = -1; + gf_boolean_t is_force_blocker = -1; gf_boolean_t exists = _gf_false; glusterd_conf_t *conf = NULL; glusterd_volinfo_t *volinfo = NULL; @@ -1709,6 +1875,15 @@ glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr) return -1; } + if (conf->op_version < 2) { + gf_log ("", GF_LOG_ERROR, "Op Version not supported."); + snprintf (errmsg, sizeof(errmsg), "One or more nodes do not" + " support the required op version."); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + exists = glusterd_check_volume_exists (volname); ret = glusterd_volinfo_find (volname, &volinfo); if ((ret) || (!exists)) { @@ -1721,11 +1896,11 @@ glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr) } ret = glusterd_get_slave_details_confpath (volinfo, dict, &slave_ip, - &slave_vol, &conf_path); + &slave_vol, &conf_path, + op_errstr); if (ret) { - snprintf (errmsg, sizeof (errmsg), - "Unable to fetch slave or confpath details."); - gf_log ("", GF_LOG_ERROR, "%s", errmsg); + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave or confpath details."); ret = -1; goto out; } @@ -1761,19 +1936,25 @@ glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr) volinfo->volname, slave); } + /* Checking if slave host is pingable, has proper passwordless + * ssh login setup, slave volume is created, slave vol is empty, + * and if it has enough memory and bypass in case of force if + * the error is not a force blocker */ ret = glusterd_verify_slave (volname, slave_ip, slave_vol, - op_errstr); - if ((ret) && !is_force) { - gf_log ("", GF_LOG_ERROR, - "%s is not a valid slave volume. Error: %s", - slave, *op_errstr); - ret = -1; - goto out; - } else if (ret) - gf_log ("", GF_LOG_INFO, "%s is not a valid slave" - " volume. Error: %s. Force creating geo-rep" - " session.", slave, *op_errstr); - + op_errstr, &is_force_blocker); + if (ret) { + if (is_force && !is_force_blocker) { + gf_log ("", GF_LOG_INFO, "%s is not a valid slave" + " volume. Error: %s. Force creating geo-rep" + " session.", slave, *op_errstr); + } else { + gf_log ("", GF_LOG_ERROR, + "%s is not a valid slave volume. Error: %s", + slave, *op_errstr); + ret = -1; + goto out; + } + } ret = dict_get_int32 (dict, "push_pem", &is_pem_push); if (!ret && is_pem_push) { @@ -1783,6 +1964,11 @@ glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr) conf->workdir); common_pem_file[ret] = '\0'; + ret = snprintf (hook_script, sizeof(hook_script) - 1, + "%s"GLUSTERD_CREATE_HOOK_SCRIPT, + conf->workdir); + hook_script[ret] = '\0'; + ret = lstat (common_pem_file, &stbuf); if (ret) { snprintf (errmsg, sizeof (errmsg), "%s" @@ -1796,6 +1982,19 @@ glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr) goto out; } + ret = lstat (hook_script, &stbuf); + if (ret) { + snprintf (errmsg, sizeof (errmsg), + "The hook-script (%s) required " + "for push-pem is not present. " + "Please install the hook-script " + "and retry", hook_script); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + if (!S_ISREG(stbuf.st_mode)) { snprintf (errmsg, sizeof (errmsg), "%s" " required for push-pem is" @@ -1818,6 +2017,7 @@ glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr) snprintf (errmsg, sizeof (errmsg), "Please check gsync " "config file. Unable to get statefile's name"); gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; goto out; } @@ -1876,6 +2076,7 @@ glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr) char errmsg[PATH_MAX] = {0,}; dict_t *ctx = NULL; gf_boolean_t is_force = 0; + gf_boolean_t is_force_blocker = -1; gf_boolean_t is_running = _gf_false; uuid_t uuid = {0}; char uuid_str [64] = {0}; @@ -1902,10 +2103,21 @@ glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr) } ret = glusterd_op_gsync_args_get (dict, op_errstr, - &volname, &slave, NULL); + &volname, &slave, &host_uuid); if (ret) goto out; + uuid_utoa_r (MY_UUID, uuid_str); + + if (conf->op_version < 2) { + gf_log ("", GF_LOG_ERROR, "Op Version not supported."); + snprintf (errmsg, sizeof(errmsg), "One or more nodes do not" + " support the required op version."); + *op_errstr = gf_strdup (errmsg); + ret = -1; + goto out; + } + exists = glusterd_check_volume_exists (volname); ret = glusterd_volinfo_find (volname, &volinfo); if ((ret) || (!exists)) { @@ -1918,17 +2130,28 @@ glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr) } ret = glusterd_get_slave_details_confpath (volinfo, dict, &slave_ip, - &slave_vol, &conf_path); + &slave_vol, &conf_path, + op_errstr); if (ret) { - snprintf (errmsg, sizeof (errmsg), - "Unable to fetch slave or confpath details."); - gf_log ("", GF_LOG_ERROR, "%s", errmsg); + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave or confpath details."); ret = -1; goto out; } ret = glusterd_get_statefile_name (volinfo, slave, conf_path, &statefile); if (ret) { + /* Checking if slave host is pingable, has proper passwordless + * ssh login setup */ + ret = glusterd_verify_slave (volname, slave_ip, slave_vol, + op_errstr, &is_force_blocker); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "%s is not a valid slave volume. Error: %s", + slave, *op_errstr); + goto out; + } + if (!strstr(slave, "::")) snprintf (errmsg, sizeof (errmsg), "%s is not a valid slave url.", slave); @@ -1971,11 +2194,6 @@ glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr) /* Check if all peers that are a part of the volume are up or not */ if ((type == GF_GSYNC_OPTION_TYPE_DELETE) || ((type == GF_GSYNC_OPTION_TYPE_STOP) && !is_force)) { - ret = dict_get_str (dict, "host-uuid", &host_uuid); - if (ret < 0) - goto out; - - uuid_utoa_r (MY_UUID, uuid_str); if (!strcmp (uuid_str, host_uuid)) { ret = glusterd_are_vol_all_peers_up (volinfo, &conf->peers, @@ -2170,6 +2388,7 @@ glusterd_gsync_configure (glusterd_volinfo_t *volinfo, char *slave, char *subop = NULL; char *master = NULL; char *conf_path = NULL; + char *slave_ip = NULL; char *slave_vol = NULL; struct stat stbuf = {0, }; @@ -2244,6 +2463,13 @@ glusterd_gsync_configure (glusterd_volinfo_t *volinfo, char *slave, ret = lstat (op_value, &stbuf); if (ret) { + ret = dict_get_str (dict, "slave_ip", &slave_ip); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to fetch slave IP."); + goto out; + } + ret = dict_get_str (dict, "slave_vol", &slave_vol); if (ret) { gf_log ("", GF_LOG_ERROR, @@ -2252,7 +2478,7 @@ glusterd_gsync_configure (glusterd_volinfo_t *volinfo, char *slave, } ret = glusterd_create_status_file (volinfo->volname, slave, - slave_vol, + slave_ip, slave_vol, "Switching Status File"); if (ret || lstat (op_value, &stbuf)) { gf_log ("", GF_LOG_ERROR, "Unable to create %s" @@ -2673,6 +2899,15 @@ glusterd_set_gsync_confs (glusterd_volinfo_t *volinfo) if (ret) goto out; + /** + * enable ignore-pid-check blindly as it could be needed for + * cascading setups. + */ + ret = glusterd_set_gsync_knob (volinfo, VKEY_MARKER_XTIME_FORCE, + &volfile_changed); + if (ret) + goto out; + ret = glusterd_set_gsync_knob (volinfo, VKEY_CHANGELOG, &volfile_changed); if (ret) @@ -2840,7 +3075,7 @@ glusterd_get_gsync_status (dict_t *dict, char **op_errstr, dict_t *rsp_dict) } static int -glusterd_gsync_delete (glusterd_volinfo_t *volinfo, char *slave, +glusterd_gsync_delete (glusterd_volinfo_t *volinfo, char *slave, char *slave_ip, char *slave_vol, char *path_list, dict_t *dict, dict_t *resp_dict, char **op_errstr) { @@ -2853,6 +3088,7 @@ glusterd_gsync_delete (glusterd_volinfo_t *volinfo, char *slave, char *conf_path = NULL; GF_ASSERT (slave); + GF_ASSERT (slave_ip); GF_ASSERT (slave_vol); GF_ASSERT (op_errstr); GF_ASSERT (dict); @@ -2902,8 +3138,8 @@ glusterd_gsync_delete (glusterd_volinfo_t *volinfo, char *slave, } ret = snprintf (geo_rep_dir, sizeof(geo_rep_dir) - 1, - "%s/"GEOREP"/%s-%s", gl_workdir, - volinfo->volname, slave_vol); + "%s/"GEOREP"/%s_%s_%s", gl_workdir, + volinfo->volname, slave_ip, slave_vol); geo_rep_dir[ret] = '\0'; ret = rmdir (geo_rep_dir); @@ -3248,6 +3484,7 @@ glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict) dict_t *resp_dict = NULL; char *host_uuid = NULL; char *slave = NULL; + char *slave_ip = NULL; char *slave_vol = NULL; char *volname = NULL; char *path_list = NULL; @@ -3286,6 +3523,12 @@ glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict) if (ret < 0) goto out; + ret = dict_get_str (dict, "slave_ip", &slave_ip); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to fetch slave volume name."); + goto out; + } + ret = dict_get_str (dict, "slave_vol", &slave_vol); if (ret) { gf_log ("", GF_LOG_ERROR, "Unable to fetch slave volume name."); @@ -3313,6 +3556,13 @@ glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict) if (type == GF_GSYNC_OPTION_TYPE_CONFIG) { ret = glusterd_gsync_configure (volinfo, slave, path_list, dict, resp_dict, op_errstr); + + ret = dict_set_str (resp_dict, "conf_path", conf_path); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store conf_file_path."); + goto out; + } goto out; } @@ -3321,9 +3571,9 @@ glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict) if (ret && !is_force && path_list) goto out; - ret = glusterd_gsync_delete (volinfo, slave, slave_vol, - path_list, dict, resp_dict, - op_errstr); + ret = glusterd_gsync_delete (volinfo, slave, slave_ip, + slave_vol, path_list, dict, + resp_dict, op_errstr); goto out; } @@ -3369,7 +3619,8 @@ glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict) if (!ret) { ret = glusterd_create_status_file (volinfo->volname, - slave, slave_vol, "Stopped"); + slave, slave_ip, + slave_vol, "Stopped"); if (ret) { gf_log ("", GF_LOG_ERROR, "Unable to update" "state_file. Error : %s", @@ -3391,17 +3642,25 @@ out: int glusterd_get_slave_details_confpath (glusterd_volinfo_t *volinfo, dict_t *dict, char **slave_ip, char **slave_vol, - char **conf_path) + char **conf_path, char **op_errstr) { int ret = -1; char confpath[PATH_MAX] = ""; glusterd_conf_t *priv = NULL; + char *slave = NULL; GF_ASSERT (THIS); priv = THIS->private; GF_ASSERT (priv); - ret = glusterd_get_slave_info (dict, slave_ip, slave_vol); + ret = dict_get_str (dict, "slave", &slave); + if (ret || !slave) { + gf_log ("", GF_LOG_ERROR, "Unable to fetch slave from dict"); + ret = -1; + goto out; + } + + ret = glusterd_get_slave_info (slave, slave_ip, slave_vol, op_errstr); if (ret) { gf_log ("", GF_LOG_ERROR, "Unable to fetch slave details."); @@ -3409,9 +3668,24 @@ glusterd_get_slave_details_confpath (glusterd_volinfo_t *volinfo, dict_t *dict, goto out; } + ret = dict_set_str (dict, "slave_ip", *slave_ip); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store slave IP."); + goto out; + } + + ret = dict_set_str (dict, "slave_vol", *slave_vol); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store slave volume name."); + goto out; + } + ret = snprintf (confpath, sizeof(confpath) - 1, - "%s/"GEOREP"/%s-%s/gsyncd.conf", - priv->workdir, volinfo->volname, *slave_vol); + "%s/"GEOREP"/%s_%s_%s/gsyncd.conf", + priv->workdir, volinfo->volname, + *slave_ip, *slave_vol); confpath[ret] = '\0'; *conf_path = gf_strdup (confpath); if (!(*conf_path)) { @@ -3435,74 +3709,66 @@ out: } static int -glusterd_get_slave_info (dict_t *dict, char **slave_ip, char **slave_vol) +glusterd_get_slave_info (char *slave, char **slave_ip, + char **slave_vol, char **op_errstr) { - char *tmp = NULL; - char *save_ptr = NULL; - char *slave = NULL; - char **linearr = NULL; - int32_t ret = -1; - - GF_ASSERT (dict); + char *tmp = NULL; + char *save_ptr = NULL; + char **linearr = NULL; + int32_t ret = -1; + char errmsg[PATH_MAX] = ""; + + ret = glusterd_urltransform_single (slave, "normalize", + &linearr); + if (ret == -1) { + ret = snprintf (errmsg, sizeof(errmsg) - 1, + "Invalid Url: %s", slave); + errmsg[ret] = '\0'; + *op_errstr = gf_strdup (errmsg); + gf_log ("", GF_LOG_ERROR, "Failed to normalize url"); + goto out; + } - ret = dict_get_str (dict, "slave", &slave); + tmp = strtok_r (linearr[0], "/", &save_ptr); + tmp = strtok_r (NULL, "/", &save_ptr); + slave = strtok_r (tmp, ":", &save_ptr); if (slave) { - ret = glusterd_urltransform_single (slave, "normalize", - &linearr); - if (ret == -1) { - gf_log ("", GF_LOG_ERROR, "Failed to normalize url"); - goto out; - } - - tmp = strtok_r (linearr[0], "/", &save_ptr); - tmp = strtok_r (NULL, "/", &save_ptr); - slave = strtok_r (tmp, ":", &save_ptr); - if (slave) { - *slave_ip = gf_strdup (slave); - if (!*slave_ip) { - gf_log ("", GF_LOG_ERROR, - "Failed to gf_strdup"); - ret = -1; - goto out; - } - gf_log ("", GF_LOG_DEBUG, "Slave IP : %s", *slave_ip); - ret = 0; - } else { - gf_log ("", GF_LOG_ERROR, "Invalid slave name"); - goto out; - } - - slave = strtok_r (NULL, ":", &save_ptr); - if (slave) { - *slave_vol = gf_strdup (slave); - if (!*slave_vol) { - gf_log ("", GF_LOG_ERROR, - "Failed to gf_strdup"); - ret = -1; - goto out; - } - gf_log ("", GF_LOG_DEBUG, "Slave Vol : %s", *slave_vol); - ret = 0; - } else { - gf_log ("", GF_LOG_ERROR, "Invalid slave name"); + ret = glusterd_mountbroker_check (&slave, op_errstr); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Invalid slave url: %s", *op_errstr); goto out; } - ret = dict_set_str (dict, "slave_ip", *slave_ip); - if (ret) { + *slave_ip = gf_strdup (slave); + if (!*slave_ip) { gf_log ("", GF_LOG_ERROR, - "Unable to store slave IP."); + "Failed to gf_strdup"); + ret = -1; goto out; } + gf_log ("", GF_LOG_DEBUG, "Slave IP : %s", *slave_ip); + ret = 0; + } else { + gf_log ("", GF_LOG_ERROR, "Invalid slave name"); + goto out; + } - ret = dict_set_str (dict, "slave_vol", *slave_vol); - if (ret) { + slave = strtok_r (NULL, ":", &save_ptr); + if (slave) { + *slave_vol = gf_strdup (slave); + if (!*slave_vol) { gf_log ("", GF_LOG_ERROR, - "Unable to store slave volume name."); + "Failed to gf_strdup"); + ret = -1; goto out; } - } else - gf_log ("", GF_LOG_ERROR, "Unable to fetch slave from dict"); + gf_log ("", GF_LOG_DEBUG, "Slave Vol : %s", *slave_vol); + ret = 0; + } else { + gf_log ("", GF_LOG_ERROR, "Invalid slave name"); + goto out; + } out: gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); @@ -3639,28 +3905,28 @@ create_conf_file (glusterd_conf_t *conf, char *conf_path) /* pid-file */ runinit_gsyncd_setrx (&runner, conf_path); runner_add_arg (&runner, "pid-file"); - runner_argprintf (&runner, "%s/${mastervol}-${slavevol}/${eSlave}.pid", georepdir); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.pid", georepdir); runner_add_args (&runner, ".", ".", NULL); RUN_GSYNCD_CMD; /* state-file */ runinit_gsyncd_setrx (&runner, conf_path); runner_add_arg (&runner, "state-file"); - runner_argprintf (&runner, "%s/${mastervol}-${slavevol}/${eSlave}.status", georepdir); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.status", georepdir); runner_add_args (&runner, ".", ".", NULL); RUN_GSYNCD_CMD; /* state-detail-file */ runinit_gsyncd_setrx (&runner, conf_path); runner_add_arg (&runner, "state-detail-file"); - runner_argprintf (&runner, "%s/${mastervol}-${slavevol}/${eSlave}-detail.status", georepdir); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}-detail.status", georepdir); runner_add_args (&runner, ".", ".", NULL); RUN_GSYNCD_CMD; /* state-socket */ runinit_gsyncd_setrx (&runner, conf_path); runner_add_arg (&runner, "state-socket-unencoded"); - runner_argprintf (&runner, "%s/${mastervol}-${slavevol}/${eSlave}.socket", georepdir); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.socket", georepdir); runner_add_args (&runner, ".", ".", NULL); RUN_GSYNCD_CMD; @@ -3753,7 +4019,9 @@ create_conf_file (glusterd_conf_t *conf, char *conf_path) } static int -glusterd_create_essential_dir_files (glusterd_volinfo_t *volinfo, dict_t *dict, char *slave, char *slave_vol, char **op_errstr) +glusterd_create_essential_dir_files (glusterd_volinfo_t *volinfo, dict_t *dict, + char *slave, char *slave_ip, + char *slave_vol, char **op_errstr) { int ret = -1; char *conf_path = NULL; @@ -3782,8 +4050,8 @@ glusterd_create_essential_dir_files (glusterd_volinfo_t *volinfo, dict_t *dict, goto out; } - ret = snprintf (buf, sizeof(buf) - 1, "%s/"GEOREP"/%s-%s", - conf->workdir, volinfo->volname, slave_vol); + ret = snprintf (buf, sizeof(buf) - 1, "%s/"GEOREP"/%s_%s_%s", + conf->workdir, volinfo->volname, slave_ip, slave_vol); buf[ret] = '\0'; ret = mkdir_p (buf, 0777, _gf_true); if (ret) { @@ -3826,7 +4094,8 @@ glusterd_create_essential_dir_files (glusterd_volinfo_t *volinfo, dict_t *dict, " Not creating status file again."); goto out; } else { - ret = glusterd_create_status_file (volinfo->volname, slave, slave_vol, + ret = glusterd_create_status_file (volinfo->volname, slave, + slave_ip, slave_vol, "Not Started"); if (ret || lstat (statefile, &stbuf)) { snprintf (errmsg, sizeof (errmsg), "Unable to create %s" @@ -3891,24 +4160,19 @@ glusterd_op_gsync_create (dict_t *dict, char **op_errstr, dict_t *rsp_dict) goto out; } + ret = dict_get_str (dict, "slave_ip", &slave_ip); + if (ret) { + snprintf (errmsg, sizeof (errmsg), + "Unable to fetch slave IP."); + gf_log ("", GF_LOG_ERROR, "%s", errmsg); + ret = -1; + goto out; + } + is_force = dict_get_str_boolean (dict, "force", _gf_false); uuid_utoa_r (MY_UUID, uuid_str); if (!strcmp (uuid_str, host_uuid)) { - ret = dict_get_str (dict, "slave_ip", &slave_ip); - if (ret) { - snprintf (errmsg, sizeof (errmsg), - "Unable to fetch slave IP." - " No Hooks Arguments."); - gf_log ("", GF_LOG_ERROR, "%s", errmsg); - if (is_force) { - ret = 0; - goto create_essentials; - } - ret = -1; - goto out; - } - ret = dict_get_int32 (dict, "push_pem", &is_pem_push); if (!ret && is_pem_push) { gf_log ("", GF_LOG_DEBUG, "Trying to setup" @@ -3950,7 +4214,9 @@ glusterd_op_gsync_create (dict_t *dict, char **op_errstr, dict_t *rsp_dict) create_essentials: - ret = glusterd_create_essential_dir_files (volinfo, dict, slave, slave_vol, op_errstr); + ret = glusterd_create_essential_dir_files (volinfo, dict, slave, + slave_ip, slave_vol, + op_errstr); if (ret) goto out; diff --git a/xlators/mgmt/glusterd/src/glusterd-handler.c b/xlators/mgmt/glusterd/src/glusterd-handler.c index afc223669..71d076624 100644 --- a/xlators/mgmt/glusterd/src/glusterd-handler.c +++ b/xlators/mgmt/glusterd/src/glusterd-handler.c @@ -33,6 +33,7 @@ #include "glusterd-op-sm.h" #include "glusterd-utils.h" #include "glusterd-store.h" +#include "glusterd-locks.h" #include "glusterd1-xdr.h" #include "cli1-xdr.h" @@ -54,6 +55,8 @@ #include <lvm2app.h> #endif +extern uuid_t global_txn_id; + int glusterd_big_locked_notify (struct rpc_clnt *rpc, void *mydata, rpc_clnt_event_t event, void *data, rpc_clnt_notify_t notify_fn) @@ -335,7 +338,9 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo, char *volume_id_str = NULL; struct args_pack pack = {0,}; xlator_t *this = NULL; - +#ifdef HAVE_BD_XLATOR + int caps = 0; +#endif GF_ASSERT (volinfo); GF_ASSERT (volumes); @@ -360,6 +365,21 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; + /* As of now, the snap volumes are also displayed as part of + volume info command. So this change is to display whether + the volume is original volume or the snap_volume. If + displaying of snap volumes in volume info o/p is not needed + this should be removed. + */ + snprintf (key, 256, "volume%d.snap_volume", count); + ret = dict_set_int32 (volumes, key, volinfo->is_snap_volume); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "failed to set whether " + "the volume is a snap volume or actual volume (%s)", + volinfo->volname); + goto out; + } + snprintf (key, 256, "volume%d.brick_count", count); ret = dict_set_int32 (volumes, key, volinfo->brick_count); if (ret) @@ -400,10 +420,71 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo, goto out; #ifdef HAVE_BD_XLATOR - snprintf (key, 256, "volume%d.backend", count); - ret = dict_set_int32 (volumes, key, volinfo->backend); - if (ret) - goto out; + if (volinfo->caps) { + caps = 0; + snprintf (key, 256, "volume%d.xlator0", count); + buf = GF_MALLOC (256, gf_common_mt_char); + if (!buf) { + ret = ENOMEM; + goto out; + } + if (volinfo->caps & CAPS_BD) + snprintf (buf, 256, "BD"); + ret = dict_set_dynstr (volumes, key, buf); + if (ret) { + GF_FREE (buf); + goto out; + } + + if (volinfo->caps & CAPS_THIN) { + snprintf (key, 256, "volume%d.xlator0.caps%d", count, + caps++); + buf = GF_MALLOC (256, gf_common_mt_char); + if (!buf) { + ret = ENOMEM; + goto out; + } + snprintf (buf, 256, "thin"); + ret = dict_set_dynstr (volumes, key, buf); + if (ret) { + GF_FREE (buf); + goto out; + } + } + + if (volinfo->caps & CAPS_OFFLOAD_COPY) { + snprintf (key, 256, "volume%d.xlator0.caps%d", count, + caps++); + buf = GF_MALLOC (256, gf_common_mt_char); + if (!buf) { + ret = ENOMEM; + goto out; + } + snprintf (buf, 256, "offload_copy"); + ret = dict_set_dynstr (volumes, key, buf); + if (ret) { + GF_FREE (buf); + goto out; + } + } + + if (volinfo->caps & CAPS_OFFLOAD_SNAPSHOT) { + snprintf (key, 256, "volume%d.xlator0.caps%d", count, + caps++); + buf = GF_MALLOC (256, gf_common_mt_char); + if (!buf) { + ret = ENOMEM; + goto out; + } + snprintf (buf, 256, "offload_snapshot"); + ret = dict_set_dynstr (volumes, key, buf); + if (ret) { + GF_FREE (buf); + goto out; + } + } + + } #endif list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { @@ -425,6 +506,16 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; +#ifdef HAVE_BD_XLATOR + if (volinfo->caps & CAPS_BD) { + snprintf (key, 256, "volume%d.vg%d", count, i); + snprintf (brick, 1024, "%s", brickinfo->vg); + buf = gf_strdup (brick); + ret = dict_set_dynstr (volumes, key, buf); + if (ret) + goto out; + } +#endif i++; } @@ -488,10 +579,17 @@ int32_t glusterd_op_txn_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx, char *err_str, size_t err_len) { - int32_t ret = -1; - xlator_t *this = NULL; - glusterd_conf_t *priv = NULL; - int32_t locked = 0; + int32_t ret = -1; + dict_t *dict = NULL; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + int32_t locked = 0; + char *tmp = NULL; + char *volname = NULL; + uuid_t *txn_id = NULL; + uuid_t *originator_uuid = NULL; + glusterd_op_info_t txn_op_info = {{0},}; + glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; GF_ASSERT (req); GF_ASSERT ((op > GD_OP_NONE) && (op < GD_OP_MAX)); @@ -502,33 +600,140 @@ glusterd_op_txn_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx, priv = this->private; GF_ASSERT (priv); - ret = glusterd_lock (MY_UUID); + dict = ctx; + + /* Generate a transaction-id for this operation and + * save it in the dict. This transaction id distinguishes + * each transaction, and helps separate opinfos in the + * op state machine. */ + txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + if (!txn_id) + goto out; + + uuid_generate (*txn_id); + + ret = dict_set_bin (dict, "transaction_id", + txn_id, sizeof(*txn_id)); if (ret) { gf_log (this->name, GF_LOG_ERROR, - "Unable to acquire lock on localhost, ret: %d", ret); - snprintf (err_str, err_len, "Another transaction is in progress. " - "Please try again after sometime."); + "Failed to set transaction id."); goto out; } + gf_log (this->name, GF_LOG_DEBUG, + "Transaction_id = %s", uuid_utoa (*txn_id)); + + /* Save the MY_UUID as the originator_uuid. This originator_uuid + * will be used by is_origin_glusterd() to determine if a node + * is the originator node for a command. */ + originator_uuid = GF_CALLOC (1, sizeof(uuid_t), + gf_common_mt_uuid_t); + if (!originator_uuid) { + ret = -1; + goto out; + } + + uuid_copy (*originator_uuid, MY_UUID); + ret = dict_set_bin (dict, "originator_uuid", + originator_uuid, sizeof (uuid_t)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set originator uuid."); + goto out; + } + + /* Based on the op_version, acquire a cluster or mgmt_v3 lock */ + if (priv->op_version < 3) { + ret = glusterd_lock (MY_UUID); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire lock on localhost, ret: %d", + ret); + snprintf (err_str, err_len, + "Another transaction is in progress. " + "Please try again after sometime."); + goto out; + } + } else { + /* If no volname is given as a part of the command, locks will + * not be held */ + ret = dict_get_str (dict, "volname", &tmp); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Failed to get volume " + "name"); + goto local_locking_done; + } else { + /* Use a copy of volname, as cli response will be + * sent before the unlock, and the volname in the + * dict, might be removed */ + volname = gf_strdup (tmp); + if (!volname) + goto out; + } + + ret = glusterd_mgmt_v3_lock (volname, MY_UUID, "vol"); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire lock for %s", volname); + snprintf (err_str, err_len, + "Another transaction is in progress for %s. " + "Please try again after sometime.", volname); + goto out; + } + } + locked = 1; gf_log (this->name, GF_LOG_DEBUG, "Acquired lock on localhost"); - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_START_LOCK, NULL); +local_locking_done: + + /* If no volname is given as a part of the command, locks will + * not be held, hence sending stage event. */ + if (volname) + event_type = GD_OP_EVENT_START_LOCK; + else { + txn_op_info.state.state = GD_OP_STATE_LOCK_SENT; + event_type = GD_OP_EVENT_ALL_ACC; + } + + /* Save opinfo for this transaction with the transaction id */ + glusterd_txn_opinfo_init (&txn_op_info, NULL, &op, ctx, req); + + ret = glusterd_set_txn_opinfo (txn_id, &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set transaction's opinfo"); + if (ctx) + dict_unref (ctx); + goto out; + } + + ret = glusterd_op_sm_inject_event (event_type, txn_id, ctx); if (ret) { gf_log (this->name, GF_LOG_ERROR, "Failed to acquire cluster" " lock."); goto out; } - glusterd_op_set_op (op); - glusterd_op_set_ctx (ctx); - glusterd_op_set_req (req); - - out: - if (locked && ret) - glusterd_unlock (MY_UUID); + if (locked && ret) { + /* Based on the op-version, we release the + * cluster or mgmt_v3 lock */ + if (priv->op_version < 3) + glusterd_unlock (MY_UUID); + else { + ret = glusterd_mgmt_v3_unlock (volname, MY_UUID, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to release lock for %s", + volname); + ret = -1; + } + } + + if (volname) + GF_FREE (volname); gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; @@ -537,11 +742,15 @@ out: int __glusterd_handle_cluster_lock (rpcsvc_request_t *req) { - gd1_mgmt_cluster_lock_req lock_req = {{0},}; - int32_t ret = -1; - glusterd_op_lock_ctx_t *ctx = NULL; - glusterd_peerinfo_t *peerinfo = NULL; - xlator_t *this = NULL; + dict_t *op_ctx = NULL; + int32_t ret = -1; + gd1_mgmt_cluster_lock_req lock_req = {{0},}; + glusterd_op_lock_ctx_t *ctx = NULL; + glusterd_op_t op = GD_OP_EVENT_LOCK; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_op_info_t txn_op_info = {{0},}; + uuid_t *txn_id = &global_txn_id; + xlator_t *this = NULL; this = THIS; GF_ASSERT (this); @@ -576,8 +785,29 @@ __glusterd_handle_cluster_lock (rpcsvc_request_t *req) uuid_copy (ctx->uuid, lock_req.uuid); ctx->req = req; + ctx->dict = NULL; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_LOCK, ctx); + op_ctx = dict_new (); + if (!op_ctx) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set new dict"); + goto out; + } + + glusterd_txn_opinfo_init (&txn_op_info, NULL, &op, op_ctx, req); + + ret = glusterd_set_txn_opinfo (txn_id, &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set transaction's opinfo"); + dict_unref (txn_op_info.op_ctx); + goto out; + } + + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_LOCK, txn_id, ctx); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to inject event GD_OP_EVENT_LOCK"); out: gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -653,6 +883,9 @@ __glusterd_handle_stage_op (rpcsvc_request_t *req) gd1_mgmt_stage_op_req op_req = {{0},}; glusterd_peerinfo_t *peerinfo = NULL; xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; + glusterd_op_info_t txn_op_info = {{0},}; + glusterd_op_sm_state_info_t state; this = THIS; GF_ASSERT (this); @@ -681,7 +914,36 @@ __glusterd_handle_stage_op (rpcsvc_request_t *req) if (ret) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_OP, req_ctx); + ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + + /* In cases where there is no volname, the receivers won't have a + * transaction opinfo created, as for those operations, the locking + * phase where the transaction opinfos are created, won't be called. */ + ret = glusterd_get_txn_opinfo (txn_id, &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "No transaction's opinfo set"); + + state.state = GD_OP_STATE_LOCKED; + glusterd_txn_opinfo_init (&txn_op_info, &state, + &op_req.op, req_ctx->dict, req); + + ret = glusterd_set_txn_opinfo (txn_id, &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set transaction's opinfo"); + dict_unref (req_ctx->dict); + goto out; + } + } + + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_OP, + txn_id, req_ctx); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to inject event GD_OP_EVENT_STAGE_OP"); out: free (op_req.buf.buf_val);//malloced by xdr @@ -705,6 +967,7 @@ __glusterd_handle_commit_op (rpcsvc_request_t *req) gd1_mgmt_commit_op_req op_req = {{0},}; glusterd_peerinfo_t *peerinfo = NULL; xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; this = THIS; GF_ASSERT (this); @@ -735,11 +998,12 @@ __glusterd_handle_commit_op (rpcsvc_request_t *req) if (ret) goto out; - ret = glusterd_op_init_ctx (op_req.op); - if (ret) - goto out; + ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_OP, req_ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_OP, + txn_id, req_ctx); out: free (op_req.buf.buf_val);//malloced by xdr @@ -1108,79 +1372,6 @@ glusterd_handle_cli_get_volume (rpcsvc_request_t *req) __glusterd_handle_cli_get_volume); } -#ifdef HAVE_BD_XLATOR -int -__glusterd_handle_cli_bd_op (rpcsvc_request_t *req) -{ - int32_t ret = -1; - gf_cli_req cli_req = { {0,} }; - dict_t *dict = NULL; - char *volname = NULL; - char op_errstr[2048] = {0,}; - glusterd_op_t cli_op = GD_OP_BD_OP; - - GF_ASSERT (req); - - ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req); - if (ret < 0) { - /* failed to decode msg */ - req->rpc_err = GARBAGE_ARGS; - goto out; - } - - gf_log ("glusterd", GF_LOG_DEBUG, "Received bd op req"); - - if (cli_req.dict.dict_len) { - /* Unserialize the dictionary */ - dict = dict_new (); - - ret = dict_unserialize (cli_req.dict.dict_val, - cli_req.dict.dict_len, - &dict); - if (ret < 0) { - gf_log ("glusterd", GF_LOG_ERROR, - "failed to " - "unserialize req-buffer to dictionary"); - goto out; - } else { - dict->extra_stdfree = cli_req.dict.dict_val; - } - } - - ret = dict_get_str (dict, "volname", &volname); - if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, - "failed to get volname"); - goto out; - } - - ret = glusterd_op_begin (req, GD_OP_BD_OP, dict, op_errstr, - sizeof (op_errstr)); -out: - if (ret && dict) - dict_unref (dict); - - glusterd_friend_sm (); - glusterd_op_sm (); - - if (ret) { - if (op_errstr[0] == '\0') - snprintf (op_errstr, sizeof (op_errstr), - "Operation failed"); - ret = glusterd_op_send_cli_response (cli_op, ret, 0, - req, NULL, op_errstr); - } - - return ret; -} - -int -glusterd_handle_cli_bd_op (rpcsvc_request_t *req) -{ - return glusterd_big_locked_handler (req, __glusterd_handle_cli_bd_op); -} -#endif - int __glusterd_handle_cli_uuid_reset (rpcsvc_request_t *req) { @@ -1863,6 +2054,57 @@ glusterd_op_unlock_send_resp (rpcsvc_request_t *req, int32_t status) } int +glusterd_op_mgmt_v3_lock_send_resp (rpcsvc_request_t *req, uuid_t *txn_id, + int32_t status) +{ + + gd1_mgmt_v3_lock_rsp rsp = {{0},}; + int ret = -1; + + GF_ASSERT (req); + GF_ASSERT (txn_id); + glusterd_get_uuid (&rsp.uuid); + rsp.op_ret = status; + if (rsp.op_ret) + rsp.op_errno = errno; + uuid_copy (rsp.txn_id, *txn_id); + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp); + + gf_log (THIS->name, GF_LOG_DEBUG, "Responded to mgmt_v3 lock, ret: %d", + ret); + + return ret; +} + +int +glusterd_op_mgmt_v3_unlock_send_resp (rpcsvc_request_t *req, uuid_t *txn_id, + int32_t status) +{ + + gd1_mgmt_v3_unlock_rsp rsp = {{0},}; + int ret = -1; + + GF_ASSERT (req); + GF_ASSERT (txn_id); + rsp.op_ret = status; + if (rsp.op_ret) + rsp.op_errno = errno; + glusterd_get_uuid (&rsp.uuid); + uuid_copy (rsp.txn_id, *txn_id); + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp); + + gf_log (THIS->name, GF_LOG_DEBUG, + "Responded to mgmt_v3 unlock, ret: %d", + ret); + + return ret; +} + +int __glusterd_handle_cluster_unlock (rpcsvc_request_t *req) { gd1_mgmt_cluster_unlock_req unlock_req = {{0}, }; @@ -1870,6 +2112,7 @@ __glusterd_handle_cluster_unlock (rpcsvc_request_t *req) glusterd_op_lock_ctx_t *ctx = NULL; glusterd_peerinfo_t *peerinfo = NULL; xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; this = THIS; GF_ASSERT (this); @@ -1904,8 +2147,9 @@ __glusterd_handle_cluster_unlock (rpcsvc_request_t *req) } uuid_copy (ctx->uuid, unlock_req.uuid); ctx->req = req; + ctx->dict = NULL; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_UNLOCK, ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_UNLOCK, txn_id, ctx); out: glusterd_friend_sm (); @@ -3059,9 +3303,13 @@ glusterd_xfer_friend_add_resp (rpcsvc_request_t *req, char *myhostname, } static void -get_probe_error_str (int op_ret, int op_errno, char *errstr, size_t len, - char *hostname, int port) +set_probe_error_str (int op_ret, int op_errno, char *op_errstr, char *errstr, + size_t len, char *hostname, int port) { + if ((op_errstr) && (strcmp (op_errstr, ""))) { + snprintf (errstr, len, "%s", op_errstr); + return; + } if (!op_ret) { switch (op_errno) { @@ -3141,11 +3389,8 @@ glusterd_xfer_cli_probe_resp (rpcsvc_request_t *req, int32_t op_ret, GF_ASSERT (req); GF_ASSERT (this); - if (op_errstr == NULL) - (void) get_probe_error_str (op_ret, op_errno, errstr, - sizeof (errstr), hostname, port); - else - snprintf (errstr, sizeof (errstr), "%s", op_errstr); + (void) set_probe_error_str (op_ret, op_errno, op_errstr, errstr, + sizeof (errstr), hostname, port); if (dict) { ret = dict_get_str (dict, "cmd-str", &cmd_str); @@ -3174,9 +3419,14 @@ glusterd_xfer_cli_probe_resp (rpcsvc_request_t *req, int32_t op_ret, } static void -get_deprobe_error_str (int op_ret, int op_errno, char *errstr, size_t len, - char *hostname) +set_deprobe_error_str (int op_ret, int op_errno, char *op_errstr, char *errstr, + size_t len, char *hostname) { + if ((op_errstr) && (strcmp (op_errstr, ""))) { + snprintf (errstr, len, "%s", op_errstr); + return; + } + if (op_ret) { switch (op_errno) { case GF_DEPROBE_LOCALHOST: @@ -3228,8 +3478,8 @@ glusterd_xfer_cli_deprobe_resp (rpcsvc_request_t *req, int32_t op_ret, GF_ASSERT (req); - (void) get_deprobe_error_str (op_ret, op_errno, errstr, sizeof (errstr), - hostname); + (void) set_deprobe_error_str (op_ret, op_errno, op_errstr, errstr, + sizeof (errstr), hostname); if (dict) { ret = dict_get_str (dict, "cmd-str", &cmd_str); @@ -3615,8 +3865,12 @@ get_brickinfo_from_brickid (char *brickid, glusterd_brickinfo_t **brickinfo) uuid_parse (volid_str, volid); ret = glusterd_volinfo_find_by_volume_id (volid, &volinfo); - if (ret) - goto out; + if (ret) { + /* Check if it a snapshot volume */ + ret = glusterd_snap_volinfo_find_by_volume_id (volid, &volinfo); + if (ret) + goto out; + } ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, brickinfo); @@ -3790,6 +4044,7 @@ __glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata, glusterd_peerinfo_t *peerinfo = NULL; glusterd_peerctx_t *peerctx = NULL; gf_boolean_t quorum_action = _gf_false; + glusterd_volinfo_t *volinfo = NULL; peerctx = mydata; if (!peerctx) @@ -3817,6 +4072,20 @@ __glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata, gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_DISCONNECT %d", peerinfo->state.state); + if (peerinfo->connected) { + list_for_each_entry (volinfo, &conf->volumes, vol_list) { + ret = glusterd_mgmt_v3_unlock (volinfo->volname, + peerinfo->uuid, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_TRACE, + "Lock not released for %s", + volinfo->volname); + } + + ret = 0; + } + if ((peerinfo->quorum_contrib != QUORUM_DOWN) && (peerinfo->state.state == GD_FRIEND_STATE_BEFRIENDED)) { peerinfo->quorum_contrib = QUORUM_DOWN; @@ -3932,11 +4201,9 @@ rpcsvc_actor_t gd_svc_cli_actors[] = { [GLUSTER_CLI_STATEDUMP_VOLUME] = {"STATEDUMP_VOLUME", GLUSTER_CLI_STATEDUMP_VOLUME, glusterd_handle_cli_statedump_volume, NULL, 0, DRC_NA}, [GLUSTER_CLI_LIST_VOLUME] = {"LIST_VOLUME", GLUSTER_CLI_LIST_VOLUME, glusterd_handle_cli_list_volume, NULL, 0, DRC_NA}, [GLUSTER_CLI_CLRLOCKS_VOLUME] = {"CLEARLOCKS_VOLUME", GLUSTER_CLI_CLRLOCKS_VOLUME, glusterd_handle_cli_clearlocks_volume, NULL, 0, DRC_NA}, -#ifdef HAVE_BD_XLATOR - [GLUSTER_CLI_BD_OP] = {"BD_OP", GLUSTER_CLI_BD_OP, glusterd_handle_cli_bd_op, NULL, 0, DRC_NA}, -#endif - [GLUSTER_CLI_COPY_FILE] = {"COPY_FILE", GLUSTER_CLI_COPY_FILE, glusterd_handle_copy_file, NULL, 0, DRC_NA}, - [GLUSTER_CLI_SYS_EXEC] = {"SYS_EXEC", GLUSTER_CLI_SYS_EXEC, glusterd_handle_sys_exec, NULL, 0, DRC_NA}, + [GLUSTER_CLI_COPY_FILE] = {"COPY_FILE", GLUSTER_CLI_COPY_FILE, glusterd_handle_copy_file, NULL, 0, DRC_NA}, + [GLUSTER_CLI_SYS_EXEC] = {"SYS_EXEC", GLUSTER_CLI_SYS_EXEC, glusterd_handle_sys_exec, NULL, 0, DRC_NA}, + [GLUSTER_CLI_SNAP] = {"SNAP", GLUSTER_CLI_SNAP, glusterd_handle_snapshot, NULL, 0, DRC_NA}, }; struct rpcsvc_program gd_svc_cli_prog = { @@ -3947,3 +4214,24 @@ struct rpcsvc_program gd_svc_cli_prog = { .actors = gd_svc_cli_actors, .synctask = _gf_true, }; + +/* This is a minimal RPC prog, which contains only the readonly RPC procs from + * the cli rpcsvc + */ +rpcsvc_actor_t gd_svc_cli_actors_ro[] = { + [GLUSTER_CLI_LIST_FRIENDS] = { "LIST_FRIENDS", GLUSTER_CLI_LIST_FRIENDS, glusterd_handle_cli_list_friends, NULL, 0, DRC_NA}, + [GLUSTER_CLI_UUID_GET] = { "UUID_GET", GLUSTER_CLI_UUID_GET, glusterd_handle_cli_uuid_get, NULL, 0, DRC_NA}, + [GLUSTER_CLI_GET_VOLUME] = { "GET_VOLUME", GLUSTER_CLI_GET_VOLUME, glusterd_handle_cli_get_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_GETWD] = { "GETWD", GLUSTER_CLI_GETWD, glusterd_handle_getwd, NULL, 1, DRC_NA}, + [GLUSTER_CLI_STATUS_VOLUME] = {"STATUS_VOLUME", GLUSTER_CLI_STATUS_VOLUME, glusterd_handle_status_volume, NULL, 0, DRC_NA}, + [GLUSTER_CLI_LIST_VOLUME] = {"LIST_VOLUME", GLUSTER_CLI_LIST_VOLUME, glusterd_handle_cli_list_volume, NULL, 0, DRC_NA}, +}; + +struct rpcsvc_program gd_svc_cli_prog_ro = { + .progname = "GlusterD svc cli read-only", + .prognum = GLUSTER_CLI_PROGRAM, + .progver = GLUSTER_CLI_VERSION, + .numactors = GLUSTER_CLI_MAXVALUE, + .actors = gd_svc_cli_actors_ro, + .synctask = _gf_true, +}; diff --git a/xlators/mgmt/glusterd/src/glusterd-handshake.c b/xlators/mgmt/glusterd/src/glusterd-handshake.c index ab172f69a..0f0357c4c 100644 --- a/xlators/mgmt/glusterd/src/glusterd-handshake.c +++ b/xlators/mgmt/glusterd/src/glusterd-handshake.c @@ -30,26 +30,147 @@ extern struct rpc_clnt_program gd_peer_prog; extern struct rpc_clnt_program gd_mgmt_prog; +extern struct rpc_clnt_program gd_mgmt_v3_prog; +extern struct rpc_clnt_program gd_mgmt_v3_prog; + #define TRUSTED_PREFIX "trusted-" typedef ssize_t (*gfs_serialize_t) (struct iovec outmsg, void *data); +static int +get_snap_volname_and_volinfo (const char *volpath, char **volname, + glusterd_volinfo_t **volinfo) +{ + int ret = -1; + char *save_ptr = NULL; + char *str_token = NULL; + char *snapname = NULL; + char *volname_token = NULL; + char *vol = NULL; + glusterd_snap_t *snap = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (volpath); + GF_ASSERT (volinfo); + + str_token = gf_strdup (volpath); + if (NULL == str_token) { + goto out; + } + + /* Input volname will have below formats: + * /snaps/<snapname>/<volname>.<hostname> + * or + * /snaps/<snapname>/<parent-volname> + * We need to extract snapname and parent_volname */ + + /*split string by "/" */ + strtok_r (str_token, "/", &save_ptr); + snapname = strtok_r(NULL, "/", &save_ptr); + if (!snapname) { + gf_log(this->name, GF_LOG_ERROR, "Invalid path: %s", volpath); + goto out; + } + + volname_token = strtok_r(NULL, "/", &save_ptr); + if (!volname_token) { + gf_log(this->name, GF_LOG_ERROR, "Invalid path: %s", volpath); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + gf_log(this->name, GF_LOG_ERROR, "Failed to " + "fetch snap %s", snapname); + goto out; + } + + /* Find if its a parent volume name or snap volume + * name. This function will succeed if volname_token + * is a parent volname + */ + ret = glusterd_volinfo_find (volname_token, volinfo); + if (ret) { + *volname = gf_strdup (volname_token); + if (NULL == *volname) { + ret = -1; + goto out; + } + + ret = glusterd_snap_volinfo_find (volname_token, snap, + volinfo); + if (ret) { + /* Split the volume name */ + vol = strtok_r (volname_token, ".", &save_ptr); + if (!vol) { + gf_log(this->name, GF_LOG_ERROR, "Invalid " + "volname (%s)", volname_token); + goto out; + } + + ret = glusterd_snap_volinfo_find (vol, snap, volinfo); + if (ret) { + gf_log(this->name, GF_LOG_ERROR, "Failed to " + "fetch snap volume from volname (%s)", + vol); + goto out; + } + } + } else { + /*volname_token is parent volname*/ + ret = glusterd_snap_volinfo_find_from_parent_volname ( + volname_token, snap, volinfo); + if (ret) { + gf_log(this->name, GF_LOG_ERROR, "Failed to " + "fetch snap volume from parent " + "volname (%s)", volname_token); + goto out; + } + + /* Since volname_token is a parent volname we should + * get the snap volname here*/ + *volname = gf_strdup ((*volinfo)->volname); + if (NULL == *volname) { + ret = -1; + goto out; + } + } + +out: + if (ret && NULL != *volname) { + GF_FREE (*volname); + *volname = NULL; + } + return ret; +} + static size_t build_volfile_path (const char *volname, char *path, size_t path_len, char *trusted_str) { - struct stat stbuf = {0,}; - int32_t ret = -1; - glusterd_conf_t *priv = NULL; - char *vol = NULL; - char *dup_volname = NULL; - char *free_ptr = NULL; - char *tmp = NULL; - glusterd_volinfo_t *volinfo = NULL; - char *server = NULL; - - priv = THIS->private; + struct stat stbuf = {0,}; + int32_t ret = -1; + glusterd_conf_t *priv = NULL; + char *vol = NULL; + char *dup_volname = NULL; + char *free_ptr = NULL; + char *save_ptr = NULL; + char *str_token = NULL; + glusterd_volinfo_t *volinfo = NULL; + char *server = NULL; + const char *volname_ptr = NULL; + char path_prefix [PATH_MAX] = {0,}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (volname); + GF_ASSERT (path); if (strstr (volname, "gluster/")) { server = strchr (volname, '/') + 1; @@ -57,6 +178,22 @@ build_volfile_path (const char *volname, char *path, path, path_len); ret = 1; goto out; + } else if ((str_token = strstr (volname, "/snaps/"))) { + ret = get_snap_volname_and_volinfo (str_token, &dup_volname, + &volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get snap" + " volinfo from path (%s)", volname); + ret = -1; + goto out; + } + + snprintf (path_prefix, sizeof (path_prefix), "%s/snaps/%s", + priv->workdir, volinfo->snapshot->snapname); + + free_ptr = dup_volname; + volname_ptr = dup_volname; + goto gotvolinfo; } else if (volname[0] != '/') { /* Normal behavior */ dup_volname = gf_strdup (volname); @@ -67,39 +204,53 @@ build_volfile_path (const char *volname, char *path, dup_volname = gf_strdup (&volname[1]); } + if (!dup_volname) { + gf_log(THIS->name, GF_LOG_ERROR, "strdup failed"); + ret = -1; + goto out; + } free_ptr = dup_volname; + volname_ptr = volname; + + snprintf (path_prefix, sizeof (path_prefix), "%s/vols", + priv->workdir); ret = glusterd_volinfo_find (dup_volname, &volinfo); + if (ret) { /* Split the volume name */ - vol = strtok_r (dup_volname, ".", &tmp); + vol = strtok_r (dup_volname, ".", &save_ptr); if (!vol) goto out; + ret = glusterd_volinfo_find (vol, &volinfo); if (ret) goto out; } +gotvolinfo: if (!glusterd_auth_get_username (volinfo)) trusted_str = NULL; - ret = snprintf (path, path_len, "%s/vols/%s/%s.vol", - priv->workdir, volinfo->volname, volname); + ret = snprintf (path, path_len, "%s/%s/%s.vol", path_prefix, + volinfo->volname, volname_ptr); if (ret == -1) goto out; ret = stat (path, &stbuf); if ((ret == -1) && (errno == ENOENT)) { - snprintf (path, path_len, "%s/vols/%s/%s%s-fuse.vol", - priv->workdir, volinfo->volname, - (trusted_str ? trusted_str : ""), dup_volname); + snprintf (path, path_len, "%s/%s/%s%s-fuse.vol", + path_prefix, volinfo->volname, + (trusted_str ? trusted_str : ""), + dup_volname); + ret = stat (path, &stbuf); } if ((ret == -1) && (errno == ENOENT)) { - snprintf (path, path_len, "%s/vols/%s/%s-tcp.vol", - priv->workdir, volinfo->volname, volname); + snprintf (path, path_len, "%s/%s/%s-tcp.vol", + path_prefix, volinfo->volname, volname_ptr); } ret = 1; @@ -244,16 +395,22 @@ __server_getspec (rpcsvc_request_t *req) } trans = req->trans; + /* addrstr will be empty for cli socket connections */ ret = rpcsvc_transport_peername (trans, (char *)&addrstr, sizeof (addrstr)); if (ret) goto fail; - tmp = strrchr (addrstr, ':'); - *tmp = '\0'; + tmp = strrchr (addrstr, ':'); + if (tmp) + *tmp = '\0'; - /* we trust the local admin */ - if (gf_is_local_addr (addrstr)) { + /* The trusted volfiles are given to the glusterd owned process like NFS + * server, self-heal daemon etc., so that they are not inadvertently + * blocked by a auth.{allow,reject} setting. The trusted volfile is not + * meant for external users. + */ + if (strlen (addrstr) && gf_is_local_addr (addrstr)) { ret = build_volfile_path (volume, filename, sizeof (filename), @@ -408,12 +565,16 @@ gd_validate_cluster_op_version (xlator_t *this, int cluster_op_version, goto out; } - if (cluster_op_version < conf->op_version) { + /* The peer can only reduce its op-version when it doesn't have any + * volumes. Reducing op-version when it already contains volumes can + * lead to inconsistencies in the cluster + */ + if ((cluster_op_version < conf->op_version) && + !list_empty (&conf->volumes)) { gf_log (this->name, GF_LOG_ERROR, - "operating version %d is less than the currently " - "running version (%d) on the machine (as per peer " - "request from %s)", cluster_op_version, - conf->op_version, peerid); + "cannot reduce operating version to %d from current " + "version %d as volumes exist (as per peer request from " + "%s)", cluster_op_version, conf->op_version, peerid); goto out; } @@ -596,6 +757,19 @@ struct rpcsvc_program gluster_handshake_prog = { .numactors = GF_HNDSK_MAXVALUE, }; +/* A minimal RPC program just for the cli getspec command */ +rpcsvc_actor_t gluster_cli_getspec_actors[] = { + [GF_HNDSK_GETSPEC] = {"GETSPEC", GF_HNDSK_GETSPEC, server_getspec, NULL, 0, DRC_NA}, +}; + +struct rpcsvc_program gluster_cli_getspec_prog = { + .progname = "Gluster Handshake (CLI Getspec)", + .prognum = GLUSTER_HNDSK_PROGRAM, + .progver = GLUSTER_HNDSK_VERSION, + .actors = gluster_cli_getspec_actors, + .numactors = GF_HNDSK_MAXVALUE, +}; + char *glusterd_dump_proc[GF_DUMP_MAXVALUE] = { [GF_DUMP_NULL] = "NULL", @@ -733,16 +907,6 @@ gd_validate_peer_op_version (xlator_t *this, glusterd_peerinfo_t *peerinfo, goto out; } - /* If peer is already operating at a higher op_version reject it. - * Cluster cannot be moved to higher op_version to accomodate a peer. - */ - if (peer_op_version > conf->op_version) { - ret = gf_asprintf (errstr, "Peer %s is already at a higher " - "op-version", peerinfo->hostname); - ret = -1; - goto out; - } - ret = 0; out: gf_log (this->name , GF_LOG_DEBUG, "Peer %s %s", peerinfo->hostname, @@ -799,6 +963,7 @@ __glusterd_mgmt_hndsk_version_ack_cbk (struct rpc_req *req, struct iovec *iov, */ peerinfo->mgmt = &gd_mgmt_prog; peerinfo->peer = &gd_peer_prog; + peerinfo->mgmt_v3 = &gd_mgmt_v3_prog; ret = default_notify (this, GF_EVENT_CHILD_UP, NULL); @@ -1025,6 +1190,14 @@ glusterd_set_clnt_mgmt_program (glusterd_peerinfo_t *peerinfo, peerinfo->peer->progname, peerinfo->peer->prognum, peerinfo->peer->progver); } + + if (peerinfo->mgmt_v3) { + gf_log ("", GF_LOG_INFO, + "Using Program %s, Num (%d), Version (%d)", + peerinfo->mgmt_v3->progname, peerinfo->mgmt_v3->prognum, + peerinfo->mgmt_v3->progver); + } + ret = 0; out: return ret; diff --git a/xlators/mgmt/glusterd/src/glusterd-locks.c b/xlators/mgmt/glusterd/src/glusterd-locks.c new file mode 100644 index 000000000..0af2a186f --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-locks.c @@ -0,0 +1,637 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include "common-utils.h" +#include "cli1-xdr.h" +#include "xdr-generic.h" +#include "glusterd.h" +#include "glusterd-op-sm.h" +#include "glusterd-store.h" +#include "glusterd-utils.h" +#include "glusterd-volgen.h" +#include "glusterd-locks.h" +#include "run.h" +#include "syscall.h" + +#include <signal.h> + +#define GF_MAX_LOCKING_ENTITIES 2 + +/* Valid entities that the mgmt_v3 lock can hold locks upon * + * To add newer entities to be locked, we can just add more * + * entries to this table along with the type and default value */ +valid_entities valid_types[] = { + { "vol", _gf_true }, + { "snap", _gf_false }, + { NULL }, +}; + +static dict_t *mgmt_v3_lock; + +/* Checks if the lock request is for a valid entity */ +gf_boolean_t +glusterd_mgmt_v3_is_type_valid (char *type) +{ + int32_t i = 0; + gf_boolean_t ret = _gf_false; + + GF_ASSERT (type); + + for (i = 0; valid_types[i].type; i++) { + if (!strcmp (type, valid_types[i].type)) { + ret = _gf_true; + break; + } + } + + return ret; +} + +/* Initialize the global mgmt_v3 lock list(dict) when + * glusterd is spawned */ +int32_t +glusterd_mgmt_v3_lock_init () +{ + int32_t ret = -1; + + mgmt_v3_lock = dict_new (); + if (!mgmt_v3_lock) { + ret = -1; + goto out; + } + + ret = 0; +out: + return ret; +} + +/* Destroy the global mgmt_v3 lock list(dict) when + * glusterd cleanup is performed */ +void +glusterd_mgmt_v3_lock_fini () +{ + if (mgmt_v3_lock) + dict_destroy (mgmt_v3_lock); +} + +int32_t +glusterd_get_mgmt_v3_lock_owner (char *key, uuid_t *uuid) +{ + int32_t ret = -1; + mgmt_v3_lock_obj *lock_obj = NULL; + uuid_t no_owner = {"\0"}; + xlator_t *this = NULL; + + GF_ASSERT(THIS); + this = THIS; + + if (!key || !uuid) { + gf_log (this->name, GF_LOG_ERROR, "key or uuid is null."); + ret = -1; + goto out; + } + + ret = dict_get_bin(mgmt_v3_lock, key, (void **) &lock_obj); + if (!ret) + uuid_copy (*uuid, lock_obj->lock_owner); + else + uuid_copy (*uuid, no_owner); + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* This function is called with the locked_count and type, to * + * release all the acquired locks. */ +static int32_t +glusterd_release_multiple_locks_per_entity (dict_t *dict, uuid_t uuid, + int32_t locked_count, + char *type) +{ + char name_buf[PATH_MAX] = ""; + char *name = NULL; + int32_t i = -1; + int32_t op_ret = 0; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT (dict); + GF_ASSERT (type); + + if (locked_count == 0) { + gf_log (this->name, GF_LOG_DEBUG, + "No %s locked as part of this transaction", + type); + goto out; + } + + /* Release all the locks held */ + for (i = 0; i < locked_count; i++) { + snprintf (name_buf, sizeof(name_buf), + "%sname%d", type, i+1); + + /* Looking for volname1, volname2 or snapname1, * + * as key in the dict snapname2 */ + ret = dict_get_str (dict, name_buf, &name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get %s locked_count = %d", + name_buf, locked_count); + op_ret = ret; + continue; + } + + ret = glusterd_mgmt_v3_unlock (name, uuid, type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release lock for %s.", + name); + op_ret = ret; + } + } + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", op_ret); + return op_ret; +} + +/* Given the count and type of the entity this function acquires * + * locks on multiple elements of the same entity. For example: * + * If type is "vol" this function tries to acquire locks on multiple * + * volumes */ +static int32_t +glusterd_acquire_multiple_locks_per_entity (dict_t *dict, uuid_t uuid, + int32_t count, char *type) +{ + char name_buf[PATH_MAX] = ""; + char *name = NULL; + int32_t i = -1; + int32_t ret = -1; + int32_t locked_count = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT (dict); + GF_ASSERT (type); + + /* Locking one element after other */ + for (i = 0; i < count; i++) { + snprintf (name_buf, sizeof(name_buf), + "%sname%d", type, i+1); + + /* Looking for volname1, volname2 or snapname1, * + * as key in the dict snapname2 */ + ret = dict_get_str (dict, name_buf, &name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get %s count = %d", + name_buf, count); + break; + } + + ret = glusterd_mgmt_v3_lock (name, uuid, type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire lock for %s %s " + "on behalf of %s. Reversing " + "this transaction", type, name, + uuid_utoa(uuid)); + break; + } + locked_count++; + } + + if (count == locked_count) { + /* If all locking ops went successfuly, return as success */ + ret = 0; + goto out; + } + + /* If we failed to lock one element, unlock others and return failure */ + ret = glusterd_release_multiple_locks_per_entity (dict, uuid, + locked_count, + type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release multiple %s locks", + type); + } + ret = -1; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Given the type of entity, this function figures out if it should unlock a * + * single element of multiple elements of the said entity. For example: * + * if the type is "vol", this function will accordingly unlock a single volume * + * or multiple volumes */ +static int32_t +glusterd_mgmt_v3_unlock_entity (dict_t *dict, uuid_t uuid, char *type, + gf_boolean_t default_value) +{ + char name_buf[PATH_MAX] = ""; + char *name = NULL; + int32_t count = -1; + int32_t ret = -1; + gf_boolean_t hold_locks = _gf_false; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT (dict); + GF_ASSERT (type); + + snprintf (name_buf, sizeof(name_buf), "hold_%s_locks", type); + hold_locks = dict_get_str_boolean (dict, name_buf, default_value); + + if (hold_locks == _gf_false) { + /* Locks were not held for this particular entity * + * Hence nothing to release */ + ret = 0; + goto out; + } + + /* Looking for volcount or snapcount in the dict */ + snprintf (name_buf, sizeof(name_buf), "%scount", type); + ret = dict_get_int32 (dict, name_buf, &count); + if (ret) { + /* count is not present. Only one * + * element name needs to be unlocked */ + snprintf (name_buf, sizeof(name_buf), "%sname", + type); + ret = dict_get_str (dict, name_buf, &name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch %sname", type); + goto out; + } + + ret = glusterd_mgmt_v3_unlock (name, uuid, type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release lock for %s %s " + "on behalf of %s.", type, name, + uuid_utoa(uuid)); + goto out; + } + } else { + /* Unlocking one element name after another */ + ret = glusterd_release_multiple_locks_per_entity (dict, + uuid, + count, + type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release all %s locks", type); + goto out; + } + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Given the type of entity, this function figures out if it should lock a * + * single element or multiple elements of the said entity. For example: * + * if the type is "vol", this function will accordingly lock a single volume * + * or multiple volumes */ +static int32_t +glusterd_mgmt_v3_lock_entity (dict_t *dict, uuid_t uuid, char *type, + gf_boolean_t default_value) +{ + char name_buf[PATH_MAX] = ""; + char *name = NULL; + int32_t count = -1; + int32_t ret = -1; + gf_boolean_t hold_locks = _gf_false; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT (dict); + GF_ASSERT (type); + + snprintf (name_buf, sizeof(name_buf), "hold_%s_locks", type); + hold_locks = dict_get_str_boolean (dict, name_buf, default_value); + + if (hold_locks == _gf_false) { + /* Not holding locks for this particular entity */ + ret = 0; + goto out; + } + + /* Looking for volcount or snapcount in the dict */ + snprintf (name_buf, sizeof(name_buf), "%scount", type); + ret = dict_get_int32 (dict, name_buf, &count); + if (ret) { + /* count is not present. Only one * + * element name needs to be locked */ + snprintf (name_buf, sizeof(name_buf), "%sname", + type); + ret = dict_get_str (dict, name_buf, &name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch %sname", type); + goto out; + } + + ret = glusterd_mgmt_v3_lock (name, uuid, type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire lock for %s %s " + "on behalf of %s.", type, name, + uuid_utoa(uuid)); + goto out; + } + } else { + /* Locking one element name after another */ + ret = glusterd_acquire_multiple_locks_per_entity (dict, + uuid, + count, + type); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire all %s locks", type); + goto out; + } + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Try to release locks of multiple entities like * + * volume, snaps etc. */ +int32_t +glusterd_multiple_mgmt_v3_unlock (dict_t *dict, uuid_t uuid) +{ + int32_t i = -1; + int32_t ret = -1; + int32_t op_ret = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + + if (!dict) { + gf_log (this->name, GF_LOG_ERROR, "dict is null."); + ret = -1; + goto out; + } + + for (i = 0; valid_types[i].type; i++) { + ret = glusterd_mgmt_v3_unlock_entity + (dict, uuid, + valid_types[i].type, + valid_types[i].default_value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to unlock all %s", + valid_types[i].type); + op_ret = ret; + } + } + + ret = op_ret; +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +/* Try to acquire locks on multiple entities like * + * volume, snaps etc. */ +int32_t +glusterd_multiple_mgmt_v3_lock (dict_t *dict, uuid_t uuid) +{ + int32_t i = -1; + int32_t ret = -1; + int32_t locked_count = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + + if (!dict) { + gf_log (this->name, GF_LOG_ERROR, "dict is null."); + ret = -1; + goto out; + } + + /* Locking one entity after other */ + for (i = 0; valid_types[i].type; i++) { + ret = glusterd_mgmt_v3_lock_entity + (dict, uuid, + valid_types[i].type, + valid_types[i].default_value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to lock all %s", + valid_types[i].type); + break; + } + locked_count++; + } + + if (locked_count == GF_MAX_LOCKING_ENTITIES) { + /* If all locking ops went successfuly, return as success */ + ret = 0; + goto out; + } + + /* If we failed to lock one entity, unlock others and return failure */ + for (i = 0; i < locked_count; i++) { + ret = glusterd_mgmt_v3_unlock_entity + (dict, uuid, + valid_types[i].type, + valid_types[i].default_value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to unlock all %s", + valid_types[i].type); + } + } + ret = -1; +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_mgmt_v3_lock (const char *name, uuid_t uuid, char *type) +{ + char key[PATH_MAX] = ""; + int32_t ret = -1; + mgmt_v3_lock_obj *lock_obj = NULL; + gf_boolean_t is_valid = _gf_true; + uuid_t owner = {0}; + xlator_t *this = NULL; + + GF_ASSERT(THIS); + this = THIS; + + if (!name || !type) { + gf_log (this->name, GF_LOG_ERROR, "name or type is null."); + ret = -1; + goto out; + } + + is_valid = glusterd_mgmt_v3_is_type_valid (type); + if (is_valid != _gf_true) { + gf_log (this->name, GF_LOG_ERROR, + "Invalid entity. Cannot perform locking " + "operation on %s types", type); + ret = -1; + goto out; + } + + ret = snprintf (key, sizeof(key), "%s_%s", name, type); + if (ret != strlen(name) + 1 + strlen(type)) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Unable to create key"); + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, + "Trying to acquire lock of %s %s for %s as %s", + type, name, uuid_utoa (uuid), key); + + ret = glusterd_get_mgmt_v3_lock_owner (key, &owner); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Unable to get mgmt_v3 lock owner"); + goto out; + } + + /* If the lock has already been held for the given volume + * we fail */ + if (!uuid_is_null (owner)) { + gf_log (this->name, GF_LOG_ERROR, "Lock for %s held by %s", + name, uuid_utoa (owner)); + ret = -1; + goto out; + } + + lock_obj = GF_CALLOC (1, sizeof(mgmt_v3_lock_obj), + gf_common_mt_mgmt_v3_lock_obj_t); + if (!lock_obj) { + ret = -1; + goto out; + } + + uuid_copy (lock_obj->lock_owner, uuid); + + ret = dict_set_bin (mgmt_v3_lock, key, lock_obj, + sizeof(*lock_obj)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set lock owner in mgmt_v3 lock"); + if (lock_obj) + GF_FREE (lock_obj); + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, + "Lock for %s %s successfully held by %s", + type, name, uuid_utoa (uuid)); + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_mgmt_v3_unlock (const char *name, uuid_t uuid, char *type) +{ + char key[PATH_MAX] = ""; + int32_t ret = -1; + gf_boolean_t is_valid = _gf_true; + uuid_t owner = {0}; + xlator_t *this = NULL; + + GF_ASSERT(THIS); + this = THIS; + + if (!name || !type) { + gf_log (this->name, GF_LOG_ERROR, "name is null."); + ret = -1; + goto out; + } + + is_valid = glusterd_mgmt_v3_is_type_valid (type); + if (is_valid != _gf_true) { + gf_log (this->name, GF_LOG_ERROR, + "Invalid entity. Cannot perform unlocking " + "operation on %s types", type); + ret = -1; + goto out; + } + + ret = snprintf (key, sizeof(key), "%s_%s", + name, type); + if (ret != strlen(name) + 1 + strlen(type)) { + gf_log (this->name, GF_LOG_ERROR, "Unable to create key"); + ret = -1; + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, + "Trying to release lock of %s %s for %s as %s", + type, name, uuid_utoa (uuid), key); + + ret = glusterd_get_mgmt_v3_lock_owner (key, &owner); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Unable to get mgmt_v3 lock owner"); + goto out; + } + + if (uuid_is_null (owner)) { + gf_log (this->name, GF_LOG_ERROR, + "Lock for %s %s not held", type, name); + ret = -1; + goto out; + } + + ret = uuid_compare (uuid, owner); + if (ret) { + + gf_log (this->name, GF_LOG_ERROR, "Lock owner mismatch. " + "Lock for %s %s held by %s", + type, name, uuid_utoa (owner)); + goto out; + } + + /* Removing the mgmt_v3 lock from the global list */ + dict_del (mgmt_v3_lock, key); + + gf_log (this->name, GF_LOG_DEBUG, + "Lock for %s %s successfully released", + type, name); + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-locks.h b/xlators/mgmt/glusterd/src/glusterd-locks.h new file mode 100644 index 000000000..83eb8c997 --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-locks.h @@ -0,0 +1,51 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +#ifndef _GLUSTERD_LOCKS_H_ +#define _GLUSTERD_LOCKS_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +typedef struct mgmt_v3_lock_object_ { + uuid_t lock_owner; +} mgmt_v3_lock_obj; + +typedef struct mgmt_v3_lock_valid_entities { + char *type; /* Entity type like vol, snap */ + gf_boolean_t default_value; /* The default value that * + * determines if the locks * + * should be held for that * + * entity */ +} valid_entities; + +int32_t +glusterd_mgmt_v3_lock_init (); + +void +glusterd_mgmt_v3_lock_fini (); + +int32_t +glusterd_get_mgmt_v3_lock_owner (char *volname, uuid_t *uuid); + +int32_t +glusterd_mgmt_v3_lock (const char *key, uuid_t uuid, char *type); + +int32_t +glusterd_mgmt_v3_unlock (const char *key, uuid_t uuid, char *type); + +int32_t +glusterd_multiple_mgmt_v3_lock (dict_t *dict, uuid_t uuid); + +int32_t +glusterd_multiple_mgmt_v3_unlock (dict_t *dict, uuid_t uuid); + +#endif diff --git a/xlators/mgmt/glusterd/src/glusterd-log-ops.c b/xlators/mgmt/glusterd/src/glusterd-log-ops.c index 0136cddc9..33bd95c03 100644 --- a/xlators/mgmt/glusterd/src/glusterd-log-ops.c +++ b/xlators/mgmt/glusterd/src/glusterd-log-ops.c @@ -24,7 +24,7 @@ #include <signal.h> int -glusterd_handle_log_rotate (rpcsvc_request_t *req) +__glusterd_handle_log_rotate (rpcsvc_request_t *req) { int32_t ret = -1; gf_cli_req cli_req = {{0,}}; @@ -90,6 +90,13 @@ out: return ret; } +int +glusterd_handle_log_rotate (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + __glusterd_handle_log_rotate); +} + /* op-sm */ int glusterd_op_stage_log_rotate (dict_t *dict, char **op_errstr) diff --git a/xlators/mgmt/glusterd/src/glusterd-mem-types.h b/xlators/mgmt/glusterd/src/glusterd-mem-types.h index 51057c274..e6f6a0333 100644 --- a/xlators/mgmt/glusterd/src/glusterd-mem-types.h +++ b/xlators/mgmt/glusterd/src/glusterd-mem-types.h @@ -67,7 +67,9 @@ typedef enum gf_gld_mem_types_ { gf_gld_mt_hooks_priv_t = gf_common_mt_end + 51, gf_gld_mt_mop_commit_req_t = gf_common_mt_end + 52, gf_gld_mt_int = gf_common_mt_end + 53, - gf_gld_mt_end = gf_common_mt_end + 54, + gf_gld_mt_snap_t = gf_common_mt_end + 54, + gf_gld_mt_missed_snapinfo_t = gf_common_mt_end + 55, + gf_gld_mt_end = gf_common_mt_end + 56, } gf_gld_mem_types_t; #endif diff --git a/xlators/mgmt/glusterd/src/glusterd-mgmt-handler.c b/xlators/mgmt/glusterd/src/glusterd-mgmt-handler.c new file mode 100644 index 000000000..a5f38ce9c --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-mgmt-handler.c @@ -0,0 +1,924 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +/* rpc related syncops */ +#include "rpc-clnt.h" +#include "protocol-common.h" +#include "xdr-generic.h" +#include "glusterd1-xdr.h" +#include "glusterd-syncop.h" + +#include "glusterd.h" +#include "glusterd-utils.h" +#include "glusterd-locks.h" +#include "glusterd-mgmt.h" +#include "glusterd-op-sm.h" + +static int +glusterd_mgmt_v3_null (rpcsvc_request_t *req) +{ + return 0; +} + +static int +glusterd_mgmt_v3_lock_send_resp (rpcsvc_request_t *req, int32_t status) +{ + + gd1_mgmt_v3_lock_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + if (rsp.op_ret) + rsp.op_errno = errno; + + glusterd_get_uuid (&rsp.uuid); + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp); + + gf_log (this->name, GF_LOG_DEBUG, + "Responded to mgmt_v3 lock, ret: %d", ret); + + return ret; +} + +static int +glusterd_synctasked_mgmt_v3_lock (rpcsvc_request_t *req, + gd1_mgmt_v3_lock_req *lock_req, + glusterd_op_lock_ctx_t *ctx) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (ctx); + GF_ASSERT (ctx->dict); + + /* Trying to acquire multiple mgmt_v3 locks */ + ret = glusterd_multiple_mgmt_v3_lock (ctx->dict, ctx->uuid); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire mgmt_v3 locks for %s", + uuid_utoa (ctx->uuid)); + + ret = glusterd_mgmt_v3_lock_send_resp (req, ret); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_op_state_machine_mgmt_v3_lock (rpcsvc_request_t *req, + gd1_mgmt_v3_lock_req *lock_req, + glusterd_op_lock_ctx_t *ctx) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_op_info_t txn_op_info = {{0},}; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + glusterd_txn_opinfo_init (&txn_op_info, NULL, &lock_req->op, + ctx->dict, req); + + ret = glusterd_set_txn_opinfo (&lock_req->txn_id, &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set transaction's opinfo"); + goto out; + } + + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_LOCK, + &lock_req->txn_id, ctx); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to inject event GD_OP_EVENT_LOCK"); + +out: + glusterd_friend_sm (); + glusterd_op_sm (); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_handle_mgmt_v3_lock_fn (rpcsvc_request_t *req) +{ + gd1_mgmt_v3_lock_req lock_req = {{0},}; + int32_t ret = -1; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_op_lock_ctx_t *ctx = NULL; + xlator_t *this = NULL; + gf_boolean_t is_synctasked = _gf_false; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &lock_req, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode lock " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, "Received mgmt_v3 lock req " + "from uuid: %s", uuid_utoa (lock_req.uuid)); + + if (glusterd_friend_find_by_uuid (lock_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (lock_req.uuid)); + ret = -1; + goto out; + } + + ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_op_lock_ctx_t); + if (!ctx) { + ret = -1; + goto out; + } + + uuid_copy (ctx->uuid, lock_req.uuid); + ctx->req = req; + + ctx->dict = dict_new (); + if (!ctx->dict) { + ret = -1; + goto out; + } + + ret = dict_unserialize (lock_req.dict.dict_val, + lock_req.dict.dict_len, &ctx->dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + is_synctasked = dict_get_str_boolean (ctx->dict, + "is_synctasked", _gf_false); + if (is_synctasked) + ret = glusterd_synctasked_mgmt_v3_lock (req, &lock_req, ctx); + else + ret = glusterd_op_state_machine_mgmt_v3_lock (req, &lock_req, + ctx); + +out: + + if (ret) { + if (ctx->dict) + dict_unref (ctx->dict); + if (ctx) + GF_FREE (ctx); + } + + free (lock_req.dict.dict_val); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_mgmt_v3_pre_validate_send_resp (rpcsvc_request_t *req, + int32_t op, int32_t status, + char *op_errstr, dict_t *rsp_dict) +{ + gd1_mgmt_v3_pre_val_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + glusterd_get_uuid (&rsp.uuid); + rsp.op = op; + if (op_errstr) + rsp.op_errstr = op_errstr; + else + rsp.op_errstr = ""; + + ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get serialized length of dict"); + goto out; + } + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_pre_val_rsp); + + GF_FREE (rsp.dict.dict_val); +out: + gf_log (this->name, GF_LOG_DEBUG, + "Responded to pre validation, ret: %d", ret); + return ret; +} + +static int +glusterd_handle_pre_validate_fn (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_v3_pre_val_req op_req = {{0},}; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + char *op_errstr = NULL; + dict_t *dict = NULL; + dict_t *rsp_dict = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &op_req, + (xdrproc_t)xdr_gd1_mgmt_v3_pre_val_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to decode pre validation " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (op_req.uuid)); + ret = -1; + goto out; + } + + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (op_req.dict.dict_val, + op_req.dict.dict_len, &dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get new dictionary"); + return -1; + } + + ret = gd_mgmt_v3_pre_validate_fn (op_req.op, dict, &op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Pre Validation failed on operation %s", + gd_op_list[op_req.op]); + } + + ret = glusterd_mgmt_v3_pre_validate_send_resp (req, op_req.op, + ret, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to send Pre Validation " + "response for operation %s", + gd_op_list[op_req.op]); + goto out; + } + +out: + if (op_errstr && (strcmp (op_errstr, ""))) + GF_FREE (op_errstr); + + free (op_req.dict.dict_val); + + if (dict) + dict_unref (dict); + + if (rsp_dict) + dict_unref (rsp_dict); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_mgmt_v3_brick_op_send_resp (rpcsvc_request_t *req, + int32_t op, int32_t status, + char *op_errstr, dict_t *rsp_dict) +{ + gd1_mgmt_v3_brick_op_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + glusterd_get_uuid (&rsp.uuid); + rsp.op = op; + if (op_errstr) + rsp.op_errstr = op_errstr; + else + rsp.op_errstr = ""; + + ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get serialized length of dict"); + goto out; + } + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_brick_op_rsp); + + GF_FREE (rsp.dict.dict_val); +out: + gf_log (this->name, GF_LOG_DEBUG, + "Responded to brick op, ret: %d", ret); + return ret; +} + +static int +glusterd_handle_brick_op_fn (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_v3_brick_op_req op_req = {{0},}; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + char *op_errstr = NULL; + dict_t *dict = NULL; + dict_t *rsp_dict = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &op_req, + (xdrproc_t)xdr_gd1_mgmt_v3_brick_op_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode brick op " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (op_req.uuid)); + ret = -1; + goto out; + } + + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (op_req.dict.dict_val, + op_req.dict.dict_len, &dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get new dictionary"); + return -1; + } + + ret = gd_mgmt_v3_brick_op_fn (op_req.op, dict, &op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Brick Op failed on operation %s", + gd_op_list[op_req.op]); + } + + ret = glusterd_mgmt_v3_brick_op_send_resp (req, op_req.op, + ret, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to send brick op " + "response for operation %s", + gd_op_list[op_req.op]); + goto out; + } + +out: + if (op_errstr && (strcmp (op_errstr, ""))) + GF_FREE (op_errstr); + + free (op_req.dict.dict_val); + + if (dict) + dict_unref (dict); + + if (rsp_dict) + dict_unref (rsp_dict); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_mgmt_v3_commit_send_resp (rpcsvc_request_t *req, + int32_t op, int32_t status, + char *op_errstr, dict_t *rsp_dict) +{ + gd1_mgmt_v3_commit_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + glusterd_get_uuid (&rsp.uuid); + rsp.op = op; + if (op_errstr) + rsp.op_errstr = op_errstr; + else + rsp.op_errstr = ""; + + ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get serialized length of dict"); + goto out; + } + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_commit_rsp); + + GF_FREE (rsp.dict.dict_val); +out: + gf_log (this->name, GF_LOG_DEBUG, "Responded to commit, ret: %d", ret); + return ret; +} + +static int +glusterd_handle_commit_fn (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_v3_commit_req op_req = {{0},}; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + char *op_errstr = NULL; + dict_t *dict = NULL; + dict_t *rsp_dict = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &op_req, + (xdrproc_t)xdr_gd1_mgmt_v3_commit_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode commit " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (op_req.uuid)); + ret = -1; + goto out; + } + + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (op_req.dict.dict_val, + op_req.dict.dict_len, &dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get new dictionary"); + return -1; + } + + ret = gd_mgmt_v3_commit_fn (op_req.op, dict, &op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "commit failed on operation %s", + gd_op_list[op_req.op]); + } + + ret = glusterd_mgmt_v3_commit_send_resp (req, op_req.op, + ret, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to send commit " + "response for operation %s", + gd_op_list[op_req.op]); + goto out; + } + +out: + if (op_errstr && (strcmp (op_errstr, ""))) + GF_FREE (op_errstr); + + free (op_req.dict.dict_val); + + if (dict) + dict_unref (dict); + + if (rsp_dict) + dict_unref (rsp_dict); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_mgmt_v3_post_validate_send_resp (rpcsvc_request_t *req, + int32_t op, int32_t status, + char *op_errstr, dict_t *rsp_dict) +{ + gd1_mgmt_v3_post_val_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + glusterd_get_uuid (&rsp.uuid); + rsp.op = op; + if (op_errstr) + rsp.op_errstr = op_errstr; + else + rsp.op_errstr = ""; + + ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val, + &rsp.dict.dict_len); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get serialized length of dict"); + goto out; + } + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_post_val_rsp); + + GF_FREE (rsp.dict.dict_val); +out: + gf_log (this->name, GF_LOG_DEBUG, + "Responded to post validation, ret: %d", ret); + return ret; +} + +static int +glusterd_handle_post_validate_fn (rpcsvc_request_t *req) +{ + int32_t ret = -1; + gd1_mgmt_v3_post_val_req op_req = {{0},}; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + char *op_errstr = NULL; + dict_t *dict = NULL; + dict_t *rsp_dict = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &op_req, + (xdrproc_t)xdr_gd1_mgmt_v3_post_val_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to decode post validation " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (op_req.uuid)); + ret = -1; + goto out; + } + + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (op_req.dict.dict_val, + op_req.dict.dict_len, &dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get new dictionary"); + return -1; + } + + ret = gd_mgmt_v3_post_validate_fn (op_req.op, op_req.op_ret, dict, + &op_errstr, rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Post Validation failed on operation %s", + gd_op_list[op_req.op]); + } + + ret = glusterd_mgmt_v3_post_validate_send_resp (req, op_req.op, + ret, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to send Post Validation " + "response for operation %s", + gd_op_list[op_req.op]); + goto out; + } + +out: + if (op_errstr && (strcmp (op_errstr, ""))) + GF_FREE (op_errstr); + + free (op_req.dict.dict_val); + + if (dict) + dict_unref (dict); + + if (rsp_dict) + dict_unref (rsp_dict); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_mgmt_v3_unlock_send_resp (rpcsvc_request_t *req, int32_t status) +{ + + gd1_mgmt_v3_unlock_rsp rsp = {{0},}; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + rsp.op_ret = status; + if (rsp.op_ret) + rsp.op_errno = errno; + + glusterd_get_uuid (&rsp.uuid); + + ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp); + + gf_log (this->name, GF_LOG_DEBUG, + "Responded to mgmt_v3 unlock, ret: %d", ret); + + return ret; +} + +static int +glusterd_synctasked_mgmt_v3_unlock (rpcsvc_request_t *req, + gd1_mgmt_v3_unlock_req *unlock_req, + glusterd_op_lock_ctx_t *ctx) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (ctx); + + /* Trying to release multiple mgmt_v3 locks */ + ret = glusterd_multiple_mgmt_v3_unlock (ctx->dict, ctx->uuid); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release mgmt_v3 locks for %s", + uuid_utoa(ctx->uuid)); + } + + ret = glusterd_mgmt_v3_unlock_send_resp (req, ret); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + + +static int +glusterd_op_state_machine_mgmt_v3_unlock (rpcsvc_request_t *req, + gd1_mgmt_v3_unlock_req *lock_req, + glusterd_op_lock_ctx_t *ctx) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_UNLOCK, + &lock_req->txn_id, ctx); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to inject event GD_OP_EVENT_UNLOCK"); + + glusterd_friend_sm (); + glusterd_op_sm (); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int +glusterd_handle_mgmt_v3_unlock_fn (rpcsvc_request_t *req) +{ + gd1_mgmt_v3_unlock_req lock_req = {{0},}; + int32_t ret = -1; + glusterd_op_lock_ctx_t *ctx = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + gf_boolean_t is_synctasked = _gf_false; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + ret = xdr_to_generic (req->msg[0], &lock_req, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_req); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "Failed to decode unlock " + "request received from peer"); + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, "Received mgmt_v3 unlock req " + "from uuid: %s", uuid_utoa (lock_req.uuid)); + + if (glusterd_friend_find_by_uuid (lock_req.uuid, &peerinfo)) { + gf_log (this->name, GF_LOG_WARNING, "%s doesn't " + "belong to the cluster. Ignoring request.", + uuid_utoa (lock_req.uuid)); + ret = -1; + goto out; + } + + ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_op_lock_ctx_t); + if (!ctx) { + ret = -1; + goto out; + } + + uuid_copy (ctx->uuid, lock_req.uuid); + ctx->req = req; + + ctx->dict = dict_new (); + if (!ctx->dict) { + ret = -1; + goto out; + } + + ret = dict_unserialize (lock_req.dict.dict_val, + lock_req.dict.dict_len, &ctx->dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "failed to unserialize the dictionary"); + goto out; + } + + is_synctasked = dict_get_str_boolean (ctx->dict, + "is_synctasked", _gf_false); + if (is_synctasked) + ret = glusterd_synctasked_mgmt_v3_unlock (req, &lock_req, ctx); + else + ret = glusterd_op_state_machine_mgmt_v3_unlock (req, &lock_req, + ctx); + +out: + + if (ret) { + if (ctx->dict) + dict_unref (ctx->dict); + if (ctx) + GF_FREE (ctx); + } + + free (lock_req.dict.dict_val); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_handle_mgmt_v3_lock (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_mgmt_v3_lock_fn); +} + +static int +glusterd_handle_pre_validate (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_pre_validate_fn); +} + +static int +glusterd_handle_brick_op (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_brick_op_fn); +} + +static int +glusterd_handle_commit (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_commit_fn); +} + +static int +glusterd_handle_post_validate (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_post_validate_fn); +} + +int +glusterd_handle_mgmt_v3_unlock (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, + glusterd_handle_mgmt_v3_unlock_fn); +} + +rpcsvc_actor_t gd_svc_mgmt_v3_actors[] = { + [GLUSTERD_MGMT_V3_NULL] = { "NULL", GLUSTERD_MGMT_V3_NULL, glusterd_mgmt_v3_null, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_LOCK] = { "MGMT_V3_LOCK", GLUSTERD_MGMT_V3_LOCK, glusterd_handle_mgmt_v3_lock, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_PRE_VALIDATE] = { "PRE_VAL", GLUSTERD_MGMT_V3_PRE_VALIDATE, glusterd_handle_pre_validate, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_BRICK_OP] = { "BRCK_OP", GLUSTERD_MGMT_V3_BRICK_OP, glusterd_handle_brick_op, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_COMMIT] = { "COMMIT", GLUSTERD_MGMT_V3_COMMIT, glusterd_handle_commit, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_POST_VALIDATE] = { "POST_VAL", GLUSTERD_MGMT_V3_POST_VALIDATE, glusterd_handle_post_validate, NULL, 0, DRC_NA}, + [GLUSTERD_MGMT_V3_UNLOCK] = { "MGMT_V3_UNLOCK", GLUSTERD_MGMT_V3_UNLOCK, glusterd_handle_mgmt_v3_unlock, NULL, 0, DRC_NA}, +}; + +struct rpcsvc_program gd_svc_mgmt_v3_prog = { + .progname = "GlusterD svc mgmt v3", + .prognum = GD_MGMT_V3_PROGRAM, + .progver = GD_MGMT_V3_VERSION, + .numactors = GLUSTERD_MGMT_V3_MAXVALUE, + .actors = gd_svc_mgmt_v3_actors, + .synctask = _gf_true, +}; diff --git a/xlators/mgmt/glusterd/src/glusterd-mgmt.c b/xlators/mgmt/glusterd/src/glusterd-mgmt.c new file mode 100644 index 000000000..d52532e54 --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-mgmt.c @@ -0,0 +1,1893 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +/* rpc related syncops */ +#include "rpc-clnt.h" +#include "protocol-common.h" +#include "xdr-generic.h" +#include "glusterd1-xdr.h" +#include "glusterd-syncop.h" + +#include "glusterd.h" +#include "glusterd-utils.h" +#include "glusterd-locks.h" +#include "glusterd-mgmt.h" +#include "glusterd-op-sm.h" + +extern struct rpc_clnt_program gd_mgmt_v3_prog; + + +static void +gd_mgmt_v3_collate_errors (struct syncargs *args, int op_ret, int op_errno, + char *op_errstr, int op_code, + glusterd_peerinfo_t *peerinfo, u_char *uuid) +{ + char *peer_str = NULL; + char err_str[PATH_MAX] = "Please check log file for details."; + char op_err[PATH_MAX] = ""; + int32_t len = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (args); + GF_ASSERT (uuid); + + if (op_ret) { + args->op_ret = op_ret; + args->op_errno = op_errno; + + if (peerinfo) + peer_str = peerinfo->hostname; + else + peer_str = uuid_utoa (uuid); + + if (op_errstr && strcmp (op_errstr, "")) { + len = snprintf (err_str, sizeof(err_str) - 1, + "Error: %s", op_errstr); + err_str[len] = '\0'; + } + + switch (op_code){ + case GLUSTERD_MGMT_V3_LOCK: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Locking failed " + "on %s. %s", peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_PRE_VALIDATE: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Pre Validation failed " + "on %s. %s", peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_BRICK_OP: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Brick ops failed " + "on %s. %s", peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_COMMIT: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Commit failed on %s. %s", + peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_POST_VALIDATE: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Post Validation failed " + "on %s. %s", peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_UNLOCK: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Unlocking failed " + "on %s. %s", peer_str, err_str); + break; + } + } + op_err[len] = '\0'; + + if (args->errstr) { + len = snprintf (err_str, sizeof(err_str) - 1, + "%s\n%s", args->errstr, + op_err); + GF_FREE (args->errstr); + args->errstr = NULL; + } else + len = snprintf (err_str, sizeof(err_str) - 1, + "%s", op_err); + err_str[len] = '\0'; + + gf_log (this->name, GF_LOG_ERROR, "%s", op_err); + args->errstr = gf_strdup (err_str); + } + + return; +} + +int32_t +gd_mgmt_v3_pre_validate_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + switch (op) { + case GD_OP_SNAP: + { + ret = glusterd_snapshot_prevalidate (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Snapshot Prevalidate Failed"); + goto out; + } + + break; + } + default: + break; + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_DEBUG, "OP = %d. Returning %d", op, ret); + return ret; +} + +int32_t +gd_mgmt_v3_brick_op_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + switch (op) { + case GD_OP_SNAP: + { + ret = glusterd_snapshot_brickop (dict, op_errstr, rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "snapshot brickop " + "failed"); + goto out; + } + break; + } + default: + break; + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "OP = %d. Returning %d", op, ret); + return ret; +} + +int32_t +gd_mgmt_v3_commit_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + switch (op) { + case GD_OP_SNAP: + { + ret = glusterd_snapshot (dict, op_errstr, rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Snapshot Commit Failed"); + goto out; + } + break; + } + default: + break; + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_DEBUG, "OP = %d. Returning %d", op, ret); + return ret; +} + +int32_t +gd_mgmt_v3_post_validate_fn (glusterd_op_t op, int32_t op_ret, dict_t *dict, + char **op_errstr, dict_t *rsp_dict) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + switch (op) { + case GD_OP_SNAP: + { + ret = glusterd_snapshot_postvalidate (dict, op_ret, + op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "postvalidate operation failed"); + goto out; + } + break; + } + default: + break; + } + + ret = 0; + +out: + gf_log (this->name, GF_LOG_TRACE, "OP = %d. Returning %d", op, ret); + return ret; +} + +int32_t +gd_mgmt_v3_lock_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_lock_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + /* Even though the lock command has failed, while collating the errors + (gd_mgmt_v3_collate_errors), args->op_ret and args->op_errno will be + used. @args is obtained from frame->local. So before checking the + status of the request and going out if its a failure, args should be + set to frame->local. Otherwise, while collating args will be NULL. + This applies to other phases such as prevalidate, brickop, commit and + postvalidate also. + */ + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_LOCK, + peerinfo, rsp.uuid); + free (rsp.dict.dict_val); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_lock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_lock_cbk_fn); +} + +int +gd_mgmt_v3_lock (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + gd1_mgmt_v3_lock_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_LOCK, + gd_mgmt_v3_lock_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_lock_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_initiate_lockdown (glusterd_conf_t *conf, glusterd_op_t op, + dict_t *dict, char **op_errstr, int npeers, + gf_boolean_t *is_acquired) +{ + char *volname = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + int32_t ret = -1; + int32_t peer_cnt = 0; + struct syncargs args = {0}; + struct list_head *peers = NULL; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (is_acquired); + + peers = &conf->xaction_peers; + + /* Trying to acquire multiple mgmt_v3 locks on local node */ + ret = glusterd_multiple_mgmt_v3_lock (dict, MY_UUID); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire mgmt_v3 locks on localhost"); + goto out; + } + + *is_acquired = _gf_true; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending mgmt_v3 lock req to other nodes in the cluster */ + gd_syncargs_init (&args, NULL); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_lock (op, dict, peerinfo, &args, + MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent lock op req for %s " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); +out: + if (ret) { + if (*op_errstr) + gf_log (this->name, GF_LOG_ERROR, "%s", + *op_errstr); + + if (volname) + ret = gf_asprintf (op_errstr, + "Another transaction is in progress " + "for %s. Please try again after " + "sometime.", volname); + else + ret = gf_asprintf (op_errstr, + "Another transaction is in progress " + "Please try again after sometime."); + + if (ret == -1) + *op_errstr = NULL; + + ret = -1; + } + + return ret; +} + +int +glusterd_pre_validate_aggr_rsp_dict (glusterd_op_t op, dict_t *aggr, dict_t *rsp) +{ + int32_t ret = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (aggr); + GF_ASSERT (rsp); + + switch (op) { + case GD_OP_SNAP: + ret = glusterd_snap_pre_validate_use_rsp_dict (aggr, rsp); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to aggregate prevalidate " + "response dictionaries."); + goto out; + } + break; + default: + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Invalid op (%s)", gd_op_list[op]); + + break; + } +out: + return ret; +} + +int32_t +gd_mgmt_v3_pre_validate_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_pre_val_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + dict_t *rsp_dict = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_pre_val_rsp); + if (ret < 0) + goto out; + + if (rsp.dict.dict_len) { + /* Unserialize the dictionary */ + rsp_dict = dict_new (); + + ret = dict_unserialize (rsp.dict.dict_val, + rsp.dict.dict_len, + &rsp_dict); + if (ret < 0) { + GF_FREE (rsp.dict.dict_val); + goto out; + } else { + rsp_dict->extra_stdfree = rsp.dict.dict_val; + } + } + + uuid_copy (args->uuid, rsp.uuid); + pthread_mutex_lock (&args->lock_dict); + { + ret = glusterd_pre_validate_aggr_rsp_dict (rsp.op, args->dict, + rsp_dict); + } + pthread_mutex_unlock (&args->lock_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response from " + " node/brick"); + if (!rsp.op_ret) + op_ret = ret; + else { + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + } + } else { + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + } + +out: + if (rsp_dict) + dict_unref (rsp_dict); + + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_PRE_VALIDATE, + peerinfo, rsp.uuid); + + if (rsp.op_errstr) + free (rsp.op_errstr); + + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_pre_validate_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_pre_validate_cbk_fn); +} + +int +gd_mgmt_v3_pre_validate_req (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + int32_t ret = -1; + gd1_mgmt_v3_pre_val_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_PRE_VALIDATE, + gd_mgmt_v3_pre_validate_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_pre_val_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_pre_validate (glusterd_conf_t *conf, glusterd_op_t op, + dict_t *req_dict, char **op_errstr, int npeers) +{ + int32_t ret = -1; + int32_t peer_cnt = 0; + dict_t *rsp_dict = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (req_dict); + GF_ASSERT (op_errstr); + + peers = &conf->xaction_peers; + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create response dictionary"); + goto out; + } + + /* Pre Validation on local node */ + ret = gd_mgmt_v3_pre_validate_fn (op, req_dict, op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Pre Validation failed for " + "operation %s on local node", + gd_op_list[op]); + + if (*op_errstr == NULL) { + ret = gf_asprintf (op_errstr, + "Pre-validation failed " + "on localhost. Please " + "check log file for details"); + if (ret == -1) + *op_errstr = NULL; + + ret = -1; + } + goto out; + } + + ret = glusterd_pre_validate_aggr_rsp_dict (op, req_dict, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response from " + " node/brick"); + goto out; + } + + dict_unref (rsp_dict); + rsp_dict = NULL; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending Pre Validation req to other nodes in the cluster */ + gd_syncargs_init (&args, req_dict); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_pre_validate_req (op, req_dict, peerinfo, &args, + MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.op_ret) { + gf_log (this->name, GF_LOG_ERROR, + "Pre Validation failed on peers"); + + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + } + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent pre valaidation req for %s " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); +out: + return ret; +} + +int +glusterd_mgmt_v3_build_payload (dict_t **req, char **op_errstr, dict_t *dict, + glusterd_op_t op) +{ + int32_t ret = -1; + dict_t *req_dict = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (op_errstr); + GF_ASSERT (dict); + + req_dict = dict_new (); + if (!req_dict) + goto out; + + switch (op) { + case GD_OP_SNAP: + dict_copy (dict, req_dict); + break; + default: + break; + } + + *req = req_dict; + ret = 0; +out: + return ret; +} + +int32_t +gd_mgmt_v3_brick_op_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_brick_op_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + /* If the operation failed, then iov can be NULL. So better check the + status of the operation and then worry about iov (if the status of + the command is success) + */ + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_brick_op_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_BRICK_OP, + peerinfo, rsp.uuid); + + if (rsp.op_errstr) + free (rsp.op_errstr); + + free (rsp.dict.dict_val); + + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_brick_op_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_brick_op_cbk_fn); +} + +int +gd_mgmt_v3_brick_op_req (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + int32_t ret = -1; + gd1_mgmt_v3_brick_op_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_BRICK_OP, + gd_mgmt_v3_brick_op_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_brick_op_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_brick_op (glusterd_conf_t *conf, glusterd_op_t op, + dict_t *req_dict, char **op_errstr, int npeers) +{ + int32_t ret = -1; + int32_t peer_cnt = 0; + dict_t *rsp_dict = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (req_dict); + GF_ASSERT (op_errstr); + + peers = &conf->xaction_peers; + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create response dictionary"); + goto out; + } + + /* Perform brick op on local node */ + ret = gd_mgmt_v3_brick_op_fn (op, req_dict, op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Brick ops failed for " + "operation %s on local node", + gd_op_list[op]); + + if (*op_errstr == NULL) { + ret = gf_asprintf (op_errstr, + "Brick ops failed " + "on localhost. Please " + "check log file for details"); + if (ret == -1) + *op_errstr = NULL; + + ret = -1; + } + goto out; + } + + dict_unref (rsp_dict); + rsp_dict = NULL; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending brick op req to other nodes in the cluster */ + gd_syncargs_init (&args, NULL); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_brick_op_req (op, req_dict, peerinfo, &args, + MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.op_ret) { + gf_log (this->name, GF_LOG_ERROR, + "Brick ops failed on peers"); + + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + } + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent brick op req for %s " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); +out: + return ret; +} + +int32_t +gd_mgmt_v3_commit_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_commit_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + dict_t *rsp_dict = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_commit_rsp); + if (ret < 0) + goto out; + + if (rsp.dict.dict_len) { + /* Unserialize the dictionary */ + rsp_dict = dict_new (); + + ret = dict_unserialize (rsp.dict.dict_val, + rsp.dict.dict_len, + &rsp_dict); + if (ret < 0) { + GF_FREE (rsp.dict.dict_val); + goto out; + } else { + rsp_dict->extra_stdfree = rsp.dict.dict_val; + } + } + + uuid_copy (args->uuid, rsp.uuid); + pthread_mutex_lock (&args->lock_dict); + { + ret = glusterd_syncop_aggr_rsp_dict (rsp.op, args->dict, + rsp_dict); + } + pthread_mutex_unlock (&args->lock_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response from " + " node/brick"); + if (!rsp.op_ret) + op_ret = ret; + else { + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + } + } else { + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + } + +out: + if (rsp_dict) + dict_unref (rsp_dict); + + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_COMMIT, + peerinfo, rsp.uuid); + + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_commit_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_commit_cbk_fn); +} + +int +gd_mgmt_v3_commit_req (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + int32_t ret = -1; + gd1_mgmt_v3_commit_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_COMMIT, + gd_mgmt_v3_commit_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_commit_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_commit (glusterd_conf_t *conf, glusterd_op_t op, + dict_t *op_ctx, dict_t *req_dict, + char **op_errstr, int npeers) +{ + int32_t ret = -1; + int32_t peer_cnt = 0; + dict_t *rsp_dict = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (op_ctx); + GF_ASSERT (req_dict); + GF_ASSERT (op_errstr); + + peers = &conf->xaction_peers; + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create response dictionary"); + goto out; + } + + /* Commit on local node */ + ret = gd_mgmt_v3_commit_fn (op, req_dict, op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Commit failed for " + "operation %s on local node", + gd_op_list[op]); + + if (*op_errstr == NULL) { + ret = gf_asprintf (op_errstr, + "Commit failed " + "on localhost. Please " + "check log file for details."); + if (ret == -1) + *op_errstr = NULL; + + ret = -1; + } + goto out; + } + + ret = glusterd_syncop_aggr_rsp_dict (op, op_ctx, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + "Failed to aggregate response from " + " node/brick"); + goto out; + } + + dict_unref (rsp_dict); + rsp_dict = NULL; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending commit req to other nodes in the cluster */ + gd_syncargs_init (&args, op_ctx); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_commit_req (op, req_dict, peerinfo, &args, + MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.op_ret) { + gf_log (this->name, GF_LOG_ERROR, + "Commit failed on peers"); + + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + } + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent commit req for %s to %d " + "peers. Returning %d", gd_op_list[op], peer_cnt, ret); +out: + return ret; +} + +int32_t +gd_mgmt_v3_post_validate_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_post_val_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_post_val_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_POST_VALIDATE, + peerinfo, rsp.uuid); + if (rsp.op_errstr) + free (rsp.op_errstr); + + free (rsp.dict.dict_val); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_post_validate_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_post_validate_cbk_fn); +} + +int +gd_mgmt_v3_post_validate_req (glusterd_op_t op, int32_t op_ret, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + int32_t ret = -1; + gd1_mgmt_v3_post_val_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + req.op_ret = op_ret; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_POST_VALIDATE, + gd_mgmt_v3_post_validate_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_post_val_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_post_validate (glusterd_conf_t *conf, glusterd_op_t op, + int32_t op_ret, dict_t *dict, dict_t *req_dict, + char **op_errstr, int npeers) +{ + int32_t ret = -1; + int32_t peer_cnt = 0; + dict_t *rsp_dict = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (dict); + GF_ASSERT (req_dict); + GF_ASSERT (op_errstr); + + peers = &conf->xaction_peers; + GF_ASSERT (peers); + + rsp_dict = dict_new (); + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create response dictionary"); + goto out; + } + + /* Copy the contents of dict like missed snaps info to req_dict */ + dict_copy (dict, req_dict); + + /* Post Validation on local node */ + ret = gd_mgmt_v3_post_validate_fn (op, op_ret, req_dict, op_errstr, + rsp_dict); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Post Validation failed for " + "operation %s on local node", + gd_op_list[op]); + + if (*op_errstr == NULL) { + ret = gf_asprintf (op_errstr, + "Post-validation failed " + "on localhost. Please check " + "log file for details"); + if (ret == -1) + *op_errstr = NULL; + + ret = -1; + } + goto out; + } + + dict_unref (rsp_dict); + rsp_dict = NULL; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending Post Validation req to other nodes in the cluster */ + gd_syncargs_init (&args, req_dict); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_post_validate_req (op, op_ret, req_dict, peerinfo, + &args, MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.op_ret) { + gf_log (this->name, GF_LOG_ERROR, + "Post Validation failed on peers"); + + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + } + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent post valaidation req for %s " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); +out: + return ret; +} + +int32_t +gd_mgmt_v3_unlock_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int32_t ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_unlock_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int32_t op_ret = -1; + int32_t op_errno = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + if (!iov) { + gf_log (this->name, GF_LOG_ERROR, "iov is NULL"); + op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; + +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_UNLOCK, + peerinfo, rsp.uuid); + free (rsp.dict.dict_val); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_mgmt_v3_unlock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_mgmt_v3_unlock_cbk_fn); +} + +int +gd_mgmt_v3_unlock (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid) +{ + int32_t ret = -1; + gd1_mgmt_v3_unlock_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_ctx); + GF_ASSERT (peerinfo); + GF_ASSERT (args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_UNLOCK, + gd_mgmt_v3_unlock_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_unlock_req); + synclock_lock (&conf->big_lock); +out: + GF_FREE (req.dict.dict_val); + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_mgmt_v3_release_peer_locks (glusterd_conf_t *conf, glusterd_op_t op, + dict_t *dict, int32_t op_ret, + char **op_errstr, int npeers, + gf_boolean_t is_acquired) +{ + int32_t ret = -1; + int32_t peer_cnt = 0; + uuid_t peer_uuid = {0}; + xlator_t *this = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + struct syncargs args = {0}; + struct list_head *peers = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (conf); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + peers = &conf->xaction_peers; + + /* If the lock has not been held during this + * transaction, do not send unlock requests */ + if (!is_acquired) + goto out; + + if (!npeers) { + ret = 0; + goto out; + } + + /* Sending mgmt_v3 unlock req to other nodes in the cluster */ + gd_syncargs_init (&args, NULL); + synctask_barrier_init((&args)); + peer_cnt = 0; + list_for_each_entry (peerinfo, peers, op_peers_list) { + gd_mgmt_v3_unlock (op, dict, peerinfo, &args, + MY_UUID, peer_uuid); + peer_cnt++; + } + gd_synctask_barrier_wait((&args), peer_cnt); + + if (args.op_ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unlock failed on peers"); + + if (!op_ret && args.errstr) + *op_errstr = gf_strdup (args.errstr); + } + + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent unlock op req for %s " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); + +out: + return ret; +} + +int32_t +glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict) +{ + int32_t ret = -1; + int32_t op_ret = -1; + int32_t npeers = 0; + dict_t *req_dict = NULL; + dict_t *tmp_dict = NULL; + glusterd_conf_t *conf = NULL; + char *op_errstr = NULL; + xlator_t *this = NULL; + gf_boolean_t is_acquired = _gf_false; + uuid_t *originator_uuid = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (dict); + conf = this->private; + GF_ASSERT (conf); + + /* Save the MY_UUID as the originator_uuid. This originator_uuid + * will be used by is_origin_glusterd() to determine if a node + * is the originator node for a command. */ + originator_uuid = GF_CALLOC (1, sizeof(uuid_t), + gf_common_mt_uuid_t); + if (!originator_uuid) { + ret = -1; + goto out; + } + + uuid_copy (*originator_uuid, MY_UUID); + ret = dict_set_bin (dict, "originator_uuid", + originator_uuid, sizeof (uuid_t)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set originator_uuid."); + goto out; + } + + /* Marking the operation as complete synctasked */ + ret = dict_set_int32 (dict, "is_synctasked", _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set synctasked flag."); + goto out; + } + + /* Use a copy at local unlock as cli response will be sent before + * the unlock and the volname in the dict might be removed */ + tmp_dict = dict_new(); + if (!tmp_dict) { + gf_log (this->name, GF_LOG_ERROR, "Unable to create dict"); + goto out; + } + dict_copy (dict, tmp_dict); + + /* BUILD PEERS LIST */ + INIT_LIST_HEAD (&conf->xaction_peers); + npeers = gd_build_peers_list (&conf->peers, &conf->xaction_peers, op); + + /* LOCKDOWN PHASE - Acquire mgmt_v3 locks */ + ret = glusterd_mgmt_v3_initiate_lockdown (conf, op, dict, &op_errstr, + npeers, &is_acquired); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "mgmt_v3 lockdown failed."); + goto out; + } + + /* BUILD PAYLOAD */ + ret = glusterd_mgmt_v3_build_payload (&req_dict, &op_errstr, dict, op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD, + gd_op_list[op]); + if (op_errstr == NULL) + gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD); + goto out; + } + + /* PRE-COMMIT VALIDATE PHASE */ + ret = glusterd_mgmt_v3_pre_validate (conf, op, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Pre Validation Failed"); + goto out; + } + + /* COMMIT OP PHASE */ + ret = glusterd_mgmt_v3_commit (conf, op, dict, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Commit Op Failed"); + goto out; + } + + /* POST-COMMIT VALIDATE PHASE */ + /* As of now, post_validate is not handling any other + commands other than snapshot. So as of now, I am + sending 0 (op_ret as 0). + */ + ret = glusterd_mgmt_v3_post_validate (conf, op, 0, dict, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Post Validation Failed"); + goto out; + } + + ret = 0; +out: + op_ret = ret; + /* UNLOCK PHASE FOR PEERS*/ + (void) glusterd_mgmt_v3_release_peer_locks (conf, op, dict, + op_ret, &op_errstr, + npeers, is_acquired); + + /* LOCAL VOLUME(S) UNLOCK */ + if (is_acquired) { + /* Trying to release multiple mgmt_v3 locks */ + ret = glusterd_multiple_mgmt_v3_unlock (tmp_dict, MY_UUID); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release mgmt_v3 locks on localhost"); + op_ret = ret; + } + } + + /* SEND CLI RESPONSE */ + glusterd_op_send_cli_response (op, op_ret, 0, req, dict, op_errstr); + + if (req_dict) + dict_unref (req_dict); + + if (tmp_dict) + dict_unref (tmp_dict); + + if (op_errstr) { + GF_FREE (op_errstr); + op_errstr = NULL; + } + + return 0; +} + +int32_t +glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict) +{ + int32_t ret = -1; + int32_t op_ret = -1; + int32_t npeers = 0; + dict_t *req_dict = NULL; + dict_t *tmp_dict = NULL; + glusterd_conf_t *conf = NULL; + char *op_errstr = NULL; + xlator_t *this = NULL; + gf_boolean_t is_acquired = _gf_false; + uuid_t *originator_uuid = NULL; + gf_boolean_t success = _gf_false; + char *tmp_errstr = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (dict); + conf = this->private; + GF_ASSERT (conf); + + /* Save the MY_UUID as the originator_uuid. This originator_uuid + * will be used by is_origin_glusterd() to determine if a node + * is the originator node for a command. */ + originator_uuid = GF_CALLOC (1, sizeof(uuid_t), + gf_common_mt_uuid_t); + if (!originator_uuid) { + ret = -1; + goto out; + } + + uuid_copy (*originator_uuid, MY_UUID); + ret = dict_set_bin (dict, "originator_uuid", + originator_uuid, sizeof (uuid_t)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set originator_uuid."); + goto out; + } + + /* Marking the operation as complete synctasked */ + ret = dict_set_int32 (dict, "is_synctasked", _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set synctasked flag."); + goto out; + } + + /* Use a copy at local unlock as cli response will be sent before + * the unlock and the volname in the dict might be removed */ + tmp_dict = dict_new(); + if (!tmp_dict) { + gf_log (this->name, GF_LOG_ERROR, "Unable to create dict"); + goto out; + } + dict_copy (dict, tmp_dict); + + /* BUILD PEERS LIST */ + INIT_LIST_HEAD (&conf->xaction_peers); + npeers = gd_build_peers_list (&conf->peers, &conf->xaction_peers, op); + + /* LOCKDOWN PHASE - Acquire mgmt_v3 locks */ + ret = glusterd_mgmt_v3_initiate_lockdown (conf, op, dict, &op_errstr, + npeers, &is_acquired); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "mgmt_v3 lockdown failed."); + goto out; + } + + /* BUILD PAYLOAD */ + ret = glusterd_mgmt_v3_build_payload (&req_dict, &op_errstr, dict, op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD, + gd_op_list[op]); + if (op_errstr == NULL) + gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD); + goto out; + } + + /* PRE-COMMIT VALIDATE PHASE */ + ret = glusterd_mgmt_v3_pre_validate (conf, op, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Pre Validation Failed"); + goto out; + } + + /* BRICK OP PHASE for initiating barrier*/ + ret = dict_set_int32 (req_dict, "barrier", 1); + if (ret) + goto out; + ret = glusterd_mgmt_v3_brick_op (conf, op, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Brick Ops Failed"); + goto unbarrier; + } + + /* COMMIT OP PHASE */ + /* TODO: As of now, the plan is to do quorum check before sending the + commit fop and if the quorum succeeds, then commit is sent to all + the other glusterds. + snap create functionality now creates the in memory and on disk + objects for the snapshot (marking them as incomplete), takes the lvm + snapshot and then updates the status of the in memory and on disk + snap objects as complete. Suppose one of the glusterds goes down + after taking the lvm snapshot, but before updating the snap object, + then treat it as a snapshot create failure and trigger cleanup. + i.e the number of commit responses received by the originator + glusterd shold be the same as the number of peers it has sent the + request to (i.e npeers variable). If not, then originator glusterd + will initiate cleanup in post-validate fop. + Question: What if one of the other glusterds goes down as explained + above and along with it the originator glusterd also goes down? + Who will initiate the cleanup? + */ + ret = glusterd_mgmt_v3_commit (conf, op, dict, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Commit Op Failed"); + /* If the main op fails, we should save the error string. + Because, op_errstr will be used for unbarrier and + unlock ops also. We might lose the actual error that + caused the failure. + */ + tmp_errstr = op_errstr; + op_errstr = NULL; + goto unbarrier; + } + + success = _gf_true; +unbarrier: + /* BRICK OP PHASE for removing the barrier*/ + ret = dict_set_int32 (req_dict, "barrier", 0); + if (ret) + goto out; + ret = glusterd_mgmt_v3_brick_op (conf, op, req_dict, + &op_errstr, npeers); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Brick Ops Failed"); + goto out; + } + + ret = 0; + +out: + op_ret = ret; + + if (success == _gf_false) + op_ret = -1; + + /* POST-COMMIT VALIDATE PHASE */ + ret = glusterd_mgmt_v3_post_validate (conf, op, op_ret, dict, req_dict, + &op_errstr, npeers); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Post Validation Failed"); + op_ret = -1; + } + + /* UNLOCK PHASE FOR PEERS*/ + (void) glusterd_mgmt_v3_release_peer_locks (conf, op, dict, + op_ret, &op_errstr, + npeers, is_acquired); + + /* If the commit op (snapshot taking) failed, then the error is stored + in tmp_errstr and unbarrier is called. Suppose, if unbarrier also + fails, then the error happened in unbarrier is logged and freed. + The error happened in commit op, which is stored in tmp_errstr + is sent to cli. + */ + if (tmp_errstr) { + if (op_errstr) { + gf_log (this->name, GF_LOG_ERROR, "unbarrier brick op" + "failed with the error %s", op_errstr); + GF_FREE (op_errstr); + op_errstr = NULL; + } + op_errstr = tmp_errstr; + } + + /* LOCAL VOLUME(S) UNLOCK */ + if (is_acquired) { + /* Trying to release multiple mgmt_v3 locks */ + ret = glusterd_multiple_mgmt_v3_unlock (tmp_dict, MY_UUID); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to release mgmt_v3 locks on localhost"); + op_ret = ret; + } + } + + /* SEND CLI RESPONSE */ + glusterd_op_send_cli_response (op, op_ret, 0, req, dict, op_errstr); + + if (req_dict) + dict_unref (req_dict); + + if (tmp_dict) + dict_unref (tmp_dict); + + if (op_errstr) { + GF_FREE (op_errstr); + op_errstr = NULL; + } + + return 0; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-mgmt.h b/xlators/mgmt/glusterd/src/glusterd-mgmt.h new file mode 100644 index 000000000..b185a9bec --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-mgmt.h @@ -0,0 +1,45 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +#ifndef _GLUSTERD_MGMT_H_ +#define _GLUSTERD_MGMT_H_ + +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +int32_t +gd_mgmt_v3_pre_validate_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict); + +int32_t +gd_mgmt_v3_brick_op_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict); + +int32_t +gd_mgmt_v3_commit_fn (glusterd_op_t op, dict_t *dict, + char **op_errstr, dict_t *rsp_dict); + +int32_t +gd_mgmt_v3_post_validate_fn (glusterd_op_t op, int32_t op_ret, dict_t *dict, + char **op_errstr, dict_t *rsp_dict); + +int32_t +glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict); + +int32_t +glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict); + +int +glusterd_snap_pre_validate_use_rsp_dict (dict_t *dst, dict_t *src); + +#endif /* _GLUSTERD_MGMT_H_ */ diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c index f1e1e0652..1666f5e4d 100644 --- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c +++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c @@ -37,6 +37,7 @@ #include "glusterd-store.h" #include "glusterd-hooks.h" #include "glusterd-volgen.h" +#include "glusterd-locks.h" #include "syscall.h" #include "cli1-xdr.h" #include "common-utils.h" @@ -67,6 +68,165 @@ static struct list_head gd_op_sm_queue; pthread_mutex_t gd_op_sm_lock; glusterd_op_info_t opinfo = {{0},}; +uuid_t global_txn_id = {"\0"}; /* To be used in + * heterogeneous + * cluster with no + * transaction ids */ + +static dict_t *txn_opinfo; + +struct txn_opinfo_object_ { + glusterd_op_info_t opinfo; +}; +typedef struct txn_opinfo_object_ txn_opinfo_obj; + +int32_t +glusterd_txn_opinfo_dict_init () +{ + int32_t ret = -1; + + txn_opinfo = dict_new (); + if (!txn_opinfo) { + ret = -1; + goto out; + } + + ret = 0; +out: + return ret; +} + +void +glusterd_txn_opinfo_dict_fini () +{ + if (txn_opinfo) + dict_destroy (txn_opinfo); +} + +void +glusterd_txn_opinfo_init (glusterd_op_info_t *opinfo, + glusterd_op_sm_state_info_t *state, + glusterd_op_t *op, + dict_t *op_ctx, + rpcsvc_request_t *req) +{ + GF_ASSERT (opinfo); + + if (state) + opinfo->state = *state; + + if (op) + opinfo->op = *op; + + opinfo->op_ctx = dict_ref(op_ctx); + + if (req) + opinfo->req = req; + + return; +} + +int32_t +glusterd_get_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t *opinfo) +{ + int32_t ret = -1; + txn_opinfo_obj *opinfo_obj = NULL; + + if (!txn_id || !opinfo) { + gf_log ("", GF_LOG_ERROR, + "Empty transaction id or opinfo received."); + ret = -1; + goto out; + } + + ret = dict_get_bin(txn_opinfo, uuid_utoa (*txn_id), + (void **) &opinfo_obj); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get transaction opinfo"); + goto out; + } + + (*opinfo) = opinfo_obj->opinfo; + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_set_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t *opinfo) +{ + int32_t ret = -1; + txn_opinfo_obj *opinfo_obj = NULL; + + if (!txn_id) { + gf_log ("", GF_LOG_ERROR, "Empty transaction id received."); + ret = -1; + goto out; + } + + ret = dict_get_bin(txn_opinfo, uuid_utoa (*txn_id), + (void **) &opinfo_obj); + if (ret) { + opinfo_obj = GF_CALLOC (1, sizeof(txn_opinfo_obj), + gf_common_mt_txn_opinfo_obj_t); + if (!opinfo_obj) { + ret = -1; + goto out; + } + + ret = dict_set_bin(txn_opinfo, uuid_utoa (*txn_id), opinfo_obj, + sizeof(txn_opinfo_obj)); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set opinfo for transaction ID : %s", + uuid_utoa (*txn_id)); + goto out; + } + } + + opinfo_obj->opinfo = (*opinfo); + + ret = 0; +out: + if (ret) + if (opinfo_obj) + GF_FREE (opinfo_obj); + + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_clear_txn_opinfo (uuid_t *txn_id) +{ + int32_t ret = -1; + glusterd_op_info_t txn_op_info = {{0},}; + + if (!txn_id) { + gf_log ("", GF_LOG_ERROR, "Empty transaction id received."); + ret = -1; + goto out; + } + + ret = glusterd_get_txn_opinfo (txn_id, &txn_op_info); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Transaction opinfo not found"); + goto out; + } + + dict_unref (txn_op_info.op_ctx); + + dict_del(txn_opinfo, uuid_utoa (*txn_id)); + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + static int glusterfs_port = GLUSTERD_DEFAULT_PORT; static char *glusterd_op_sm_state_names[] = { "Default", @@ -147,10 +307,10 @@ glusterd_is_volume_started (glusterd_volinfo_t *volinfo) } static int -glusterd_op_sm_inject_all_acc () +glusterd_op_sm_inject_all_acc (uuid_t *txn_id) { int32_t ret = -1; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, txn_id, NULL); gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -235,20 +395,20 @@ glusterd_brick_op_build_payload (glusterd_op_t op, glusterd_brickinfo_t *brickin brick_req->name = gf_strdup (name); break; - -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: - { + case GD_OP_SNAP: brick_req = GF_CALLOC (1, sizeof (*brick_req), gf_gld_mt_mop_brick_req_t); if (!brick_req) goto out; - brick_req->op = GLUSTERD_BRICK_BD_OP; - brick_req->name = ""; - } + brick_req->op = GLUSTERD_VOLUME_BARRIER_OP; + ret = dict_get_str (dict, "volname", &volname); + if (ret) + goto out; + snprintf (name, 1024, "%s-server",volname); + brick_req->name = gf_strdup (name); + break; -#endif default: goto out; break; @@ -416,11 +576,17 @@ glusterd_op_stage_set_volume (dict_t *dict, char **op_errstr) if (!val_dict) goto out; + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); + goto out; + } + /* Check if we can support the required op-version * This check is not done on the originator glusterd. The originator * glusterd sets this value. */ - origin_glusterd = is_origin_glusterd (); + origin_glusterd = is_origin_glusterd (dict); if (!origin_glusterd) { /* Check for v3.3.x origin glusterd */ @@ -1101,14 +1267,17 @@ _delete_reconfig_opt (dict_t *this, char *key, data_t *value, void *data) GF_ASSERT (data); is_force = (int32_t*)data; - if (*is_force != 1 && - (_gf_true == glusterd_check_voloption_flags (key, - OPT_FLAG_FORCE))) { + if (*is_force != 1) { + if (_gf_true == glusterd_check_voloption_flags (key, + OPT_FLAG_FORCE)) { /* indicate to caller that we don't set the option * due to being protected */ - *is_force = -1; - goto out; + *is_force = *is_force | GD_OP_PROTECTED; + goto out; + } else { + *is_force = *is_force | GD_OP_UNPROTECTED; + } } gf_log ("", GF_LOG_DEBUG, "deleting dict with key=%s,value=%s", @@ -1277,7 +1446,7 @@ out: } static int -glusterd_op_reset_volume (dict_t *dict, char **op_errstr) +glusterd_op_reset_volume (dict_t *dict, char **op_rspstr) { glusterd_volinfo_t *volinfo = NULL; int ret = -1; @@ -1332,10 +1501,18 @@ glusterd_op_reset_volume (dict_t *dict, char **op_errstr) quorum_action = _gf_true; ret = glusterd_options_reset (volinfo, key, &is_force); - if (is_force == -1) { - ret = -1; - gf_asprintf(op_errstr, "'%s' is protected. To reset use 'force'.", - key); + if (ret == -1) { + gf_asprintf(op_rspstr, "Volume reset : failed"); + } else if (is_force & GD_OP_PROTECTED) { + if (is_force & GD_OP_UNPROTECTED) { + gf_asprintf (op_rspstr, "All unprotected fields were" + " reset. To reset the protected fields," + " use 'force'."); + } else { + ret = -1; + gf_asprintf (op_rspstr, "'%s' is protected. To reset" + " use 'force'.", key); + } } out: @@ -1365,14 +1542,25 @@ glusterd_stop_bricks (glusterd_volinfo_t *volinfo) int glusterd_start_bricks (glusterd_volinfo_t *volinfo) { - glusterd_brickinfo_t *brickinfo = NULL; + int ret = -1; + glusterd_brickinfo_t *brickinfo = NULL; + + GF_ASSERT (volinfo); list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { - if (glusterd_brick_start (volinfo, brickinfo, _gf_false)) - return -1; + ret = glusterd_brick_start (volinfo, brickinfo, _gf_false); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to start %s:%s for %s", + brickinfo->hostname, brickinfo->path, + volinfo->volname); + goto out; + } } - return 0; + ret = 0; +out: + return ret; } static int @@ -1475,6 +1663,7 @@ glusterd_op_set_volume (dict_t *dict) char str[50] = {0, }; char *op_errstr = NULL; gf_boolean_t global_opt = _gf_false; + gf_boolean_t global_opts_set = _gf_false; glusterd_volinfo_t *voliter = NULL; int32_t dict_count = 0; gf_boolean_t check_op_version = _gf_false; @@ -1536,7 +1725,6 @@ glusterd_op_set_volume (dict_t *dict) for (count = 1; ret != -1 ; count++) { - global_opt = _gf_false; sprintf (str, "key%d", count); ret = dict_get_str (dict, str, &key); if (ret) @@ -1584,8 +1772,11 @@ glusterd_op_set_volume (dict_t *dict) } } - if (glusterd_check_globaloption (key)) + global_opt = _gf_false; + if (glusterd_check_globaloption (key)) { global_opt = _gf_true; + global_opts_set = _gf_true; + } if (!global_opt) value = gf_strdup (value); @@ -1641,7 +1832,7 @@ glusterd_op_set_volume (dict_t *dict) } } - if (!global_opt) { + if (!global_opts_set) { gd_update_volume_op_versions (volinfo); ret = glusterd_create_volfiles_and_notify_services (volinfo); if (ret) { @@ -1901,6 +2092,105 @@ out: } static int +_add_brick_name_to_dict (dict_t *dict, char *key, glusterd_brickinfo_t *brick) +{ + int ret = -1; + char tmp[1024] = {0,}; + char *brickname = NULL; + xlator_t *this = NULL; + + GF_ASSERT (dict); + GF_ASSERT (key); + GF_ASSERT (brick); + + this = THIS; + GF_ASSERT (this); + + snprintf (tmp, sizeof (tmp), "%s:%s", brick->hostname, brick->path); + brickname = gf_strdup (tmp); + if (!brickname) { + gf_log (this->name, GF_LOG_ERROR, "Failed to dup brick name"); + goto out; + } + + ret = dict_set_dynstr (dict, key, brickname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add brick name to dict"); + goto out; + } + brickname = NULL; +out: + if (brickname) + GF_FREE (brickname); + return ret; +} + +static int +_add_remove_bricks_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, + char *prefix) +{ + int ret = -1; + int count = 0; + int i = 0; + char brick_key[1024] = {0,}; + char dict_key[1024] ={0,}; + char *brick = NULL; + xlator_t *this = NULL; + + GF_ASSERT (dict); + GF_ASSERT (volinfo); + GF_ASSERT (prefix); + + this = THIS; + GF_ASSERT (this); + + ret = dict_get_int32 (volinfo->rebal.dict, "count", &count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get brick count"); + goto out; + } + + snprintf (dict_key, sizeof (dict_key), "%s.count", prefix); + ret = dict_set_int32 (dict, dict_key, count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set brick count in dict"); + goto out; + } + + for (i = 1; i <= count; i++) { + memset (brick_key, 0, sizeof (brick_key)); + snprintf (brick_key, sizeof (brick_key), "brick%d", i); + + ret = dict_get_str (volinfo->rebal.dict, brick_key, &brick); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get %s", brick_key); + goto out; + } + + memset (dict_key, 0, sizeof (dict_key)); + snprintf (dict_key, sizeof (dict_key), "%s.%s", prefix, + brick_key); + ret = dict_set_str (dict, dict_key, brick); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add brick to dict"); + goto out; + } + brick = NULL; + } + +out: + return ret; +} + +/* This adds the respective task-id and all available parameters of a task into + * a dictionary + */ +static int _add_task_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, int op, int index) { @@ -1917,13 +2207,34 @@ _add_task_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, int op, int index) GF_ASSERT (this); switch (op) { - case GD_OP_REBALANCE: case GD_OP_REMOVE_BRICK: + snprintf (key, sizeof (key), "task%d", index); + ret = _add_remove_bricks_to_dict (dict, volinfo, key); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add remove bricks to dict"); + goto out; + } + case GD_OP_REBALANCE: uuid_str = gf_strdup (uuid_utoa (volinfo->rebal.rebalance_id)); status = volinfo->rebal.defrag_status; break; case GD_OP_REPLACE_BRICK: + snprintf (key, sizeof (key), "task%d.src-brick", index); + ret = _add_brick_name_to_dict (dict, key, + volinfo->rep_brick.src_brick); + if (ret) + goto out; + memset (key, 0, sizeof (key)); + + snprintf (key, sizeof (key), "task%d.dst-brick", index); + ret = _add_brick_name_to_dict (dict, key, + volinfo->rep_brick.dst_brick); + if (ret) + goto out; + memset (key, 0, sizeof (key)); + uuid_str = gf_strdup (uuid_utoa (volinfo->rep_brick.rb_id)); status = volinfo->rep_brick.rb_status; break; @@ -1936,8 +2247,7 @@ _add_task_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, int op, int index) } snprintf (key, sizeof (key), "task%d.type", index); - ret = dict_set_str (dict, key, - (char *)gd_op_list[op]); + ret = dict_set_str (dict, key, (char *)gd_op_list[op]); if (ret) { gf_log (this->name, GF_LOG_ERROR, "Error setting task type in dict"); @@ -1947,7 +2257,6 @@ _add_task_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, int op, int index) memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "task%d.id", index); - if (!uuid_str) goto out; ret = dict_set_dynstr (dict, key, uuid_str); @@ -1974,6 +2283,50 @@ out: } static int +glusterd_aggregate_task_status (dict_t *rsp_dict, glusterd_volinfo_t *volinfo) +{ + int ret = -1; + int tasks = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + if (!uuid_is_null (volinfo->rebal.rebalance_id)) { + ret = _add_task_to_dict (rsp_dict, volinfo, volinfo->rebal.op, + tasks); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add task details to dict"); + goto out; + } + tasks++; + } + + if (!uuid_is_null (volinfo->rep_brick.rb_id)) { + ret = _add_task_to_dict (rsp_dict, volinfo, GD_OP_REPLACE_BRICK, + tasks); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add task details to dict"); + goto out; + } + tasks++; + } + + ret = dict_set_int32 (rsp_dict, "tasks", tasks); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error setting tasks count in dict"); + goto out; + } + ret = 0; + +out: + return ret; +} + +static int glusterd_op_status_volume (dict_t *dict, char **op_errstr, dict_t *rsp_dict) { @@ -1993,7 +2346,6 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, gf_boolean_t nfs_disabled = _gf_false; gf_boolean_t shd_enabled = _gf_true; gf_boolean_t origin_glusterd = _gf_false; - int tasks = 0; this = THIS; GF_ASSERT (this); @@ -2003,13 +2355,13 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, GF_ASSERT (dict); - origin_glusterd = is_origin_glusterd (); + origin_glusterd = is_origin_glusterd (dict); ret = dict_get_uint32 (dict, "cmd", &cmd); if (ret) goto out; - if (is_origin_glusterd ()) { + if (origin_glusterd) { ret = 0; if ((cmd & GF_CLI_STATUS_ALL)) { ret = glusterd_get_all_volnames (rsp_dict); @@ -2076,6 +2428,10 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, brick_index); node_count++; + } else if ((cmd & GF_CLI_STATUS_TASKS) != 0) { + ret = glusterd_aggregate_task_status (rsp_dict, volinfo); + goto out; + } else { list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { brick_index++; @@ -2157,31 +2513,10 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, !origin_glusterd) goto out; - if (!uuid_is_null (volinfo->rebal.rebalance_id)) { - ret = _add_task_to_dict (rsp_dict, volinfo, volinfo->rebal.op, - tasks); - if (ret) { - gf_log (this->name, GF_LOG_ERROR, - "Failed to add task details to dict"); - goto out; - } - tasks++; - } - if (!uuid_is_null (volinfo->rep_brick.rb_id)) { - ret = _add_task_to_dict (rsp_dict, volinfo, GD_OP_REPLACE_BRICK, - tasks); - if (ret) { - gf_log (this->name, GF_LOG_ERROR, - "Failed to add task details to dict"); - goto out; - } - tasks++; - } - - ret = dict_set_int32 (rsp_dict, "tasks", tasks); + ret = glusterd_aggregate_task_status (rsp_dict, volinfo); if (ret) - gf_log (this->name, GF_LOG_ERROR, - "Error setting tasks count in dict"); + goto out; + ret = 0; out: gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -2208,6 +2543,7 @@ glusterd_op_ac_send_lock (glusterd_op_sm_event_t *event, void *ctx) xlator_t *this = NULL; glusterd_peerinfo_t *peerinfo = NULL; uint32_t pending_count = 0; + dict_t *dict = NULL; this = THIS; priv = this->private; @@ -2222,27 +2558,60 @@ glusterd_op_ac_send_lock (glusterd_op_sm_event_t *event, void *ctx) (glusterd_op_get_op() != GD_OP_SYNC_VOLUME)) continue; - proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_LOCK]; - if (proc->fn) { - ret = proc->fn (NULL, this, peerinfo); - if (ret) { - gf_log (this->name, GF_LOG_WARNING, "Failed to " - "send lock request for operation " - "'Volume %s' to peer %s", - gd_op_list[opinfo.op], - peerinfo->hostname); - continue; + /* Based on the op_version, acquire a cluster or mgmt_v3 lock */ + if (priv->op_version < 3) { + proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_LOCK]; + if (proc->fn) { + ret = proc->fn (NULL, this, peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send lock request " + "for operation 'Volume %s' to " + "peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + continue; + } + pending_count++; + } + } else { + dict = glusterd_op_get_ctx (); + dict_ref (dict); + + proc = &peerinfo->mgmt_v3->proctable + [GLUSTERD_MGMT_V3_LOCK]; + if (proc->fn) { + ret = dict_set_static_ptr (dict, "peerinfo", + peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to set peerinfo"); + dict_unref (dict); + goto out; + } + + ret = proc->fn (NULL, this, dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send mgmt_v3 lock " + "request for operation " + "'Volume %s' to peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + dict_unref (dict); + continue; + } + pending_count++; } - pending_count++; } } opinfo.pending_count = pending_count; if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_all_acc (); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); +out: gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); - return ret; } @@ -2255,17 +2624,12 @@ glusterd_op_ac_send_unlock (glusterd_op_sm_event_t *event, void *ctx) xlator_t *this = NULL; glusterd_peerinfo_t *peerinfo = NULL; uint32_t pending_count = 0; + dict_t *dict = NULL; this = THIS; priv = this->private; GF_ASSERT (priv); - /*ret = glusterd_unlock (MY_UUID); - - if (ret) - goto out; - */ - list_for_each_entry (peerinfo, &priv->peers, uuid_list) { GF_ASSERT (peerinfo); @@ -2275,29 +2639,62 @@ glusterd_op_ac_send_unlock (glusterd_op_sm_event_t *event, void *ctx) (glusterd_op_get_op() != GD_OP_SYNC_VOLUME)) continue; - proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_UNLOCK]; - if (proc->fn) { - ret = proc->fn (NULL, this, peerinfo); - if (ret) { - gf_log (this->name, GF_LOG_WARNING, "Failed to " - "send unlock request for operation " - "'Volume %s' to peer %s", - gd_op_list[opinfo.op], - peerinfo->hostname); - continue; + /* Based on the op_version, release the * + * cluster or mgmt_v3 lock */ + if (priv->op_version < 3) { + proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_UNLOCK]; + if (proc->fn) { + ret = proc->fn (NULL, this, peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send unlock request " + "for operation 'Volume %s' to " + "peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + continue; + } + pending_count++; + } + } else { + dict = glusterd_op_get_ctx (); + dict_ref (dict); + + proc = &peerinfo->mgmt_v3->proctable + [GLUSTERD_MGMT_V3_UNLOCK]; + if (proc->fn) { + ret = dict_set_static_ptr (dict, "peerinfo", + peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to set peerinfo"); + dict_unref (dict); + goto out; + } + + ret = proc->fn (NULL, this, dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send mgmt_v3 unlock " + "request for operation " + "'Volume %s' to peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + dict_unref (dict); + continue; + } + pending_count++; } - pending_count++; } } opinfo.pending_count = pending_count; if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_all_acc (); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); +out: gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); - return ret; - } static int @@ -2309,7 +2706,8 @@ glusterd_op_ac_ack_drain (glusterd_op_sm_event_t *event, void *ctx) opinfo.pending_count--; if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, NULL); gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret); @@ -2325,43 +2723,94 @@ glusterd_op_ac_send_unlock_drain (glusterd_op_sm_event_t *event, void *ctx) static int glusterd_op_ac_lock (glusterd_op_sm_event_t *event, void *ctx) { - glusterd_op_lock_ctx_t *lock_ctx = NULL; - int32_t ret = 0; + int32_t ret = 0; + char *volname = NULL; + glusterd_op_lock_ctx_t *lock_ctx = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; GF_ASSERT (event); GF_ASSERT (ctx); + this = THIS; + priv = this->private; + lock_ctx = (glusterd_op_lock_ctx_t *)ctx; - ret = glusterd_lock (lock_ctx->uuid); + /* If the req came from a node running on older op_version + * the dict won't be present. Based on it acquiring a cluster + * or mgmt_v3 lock */ + if (lock_ctx->dict == NULL) { + ret = glusterd_lock (lock_ctx->uuid); + glusterd_op_lock_send_resp (lock_ctx->req, ret); + } else { + ret = dict_get_str (lock_ctx->dict, "volname", &volname); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire volname"); + else { + ret = glusterd_mgmt_v3_lock (volname, lock_ctx->uuid, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire lock for %s", + volname); + } - gf_log (THIS->name, GF_LOG_DEBUG, "Lock Returned %d", ret); + glusterd_op_mgmt_v3_lock_send_resp (lock_ctx->req, + &event->txn_id, ret); - glusterd_op_lock_send_resp (lock_ctx->req, ret); + dict_unref (lock_ctx->dict); + } + gf_log (THIS->name, GF_LOG_DEBUG, "Lock Returned %d", ret); return ret; } static int glusterd_op_ac_unlock (glusterd_op_sm_event_t *event, void *ctx) { - int ret = 0; - glusterd_op_lock_ctx_t *lock_ctx = NULL; - xlator_t *this = NULL; - glusterd_conf_t *priv = NULL; + int32_t ret = 0; + char *volname = NULL; + glusterd_op_lock_ctx_t *lock_ctx = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + GF_ASSERT (event); GF_ASSERT (ctx); this = THIS; priv = this->private; + lock_ctx = (glusterd_op_lock_ctx_t *)ctx; - ret = glusterd_unlock (lock_ctx->uuid); + /* If the req came from a node running on older op_version + * the dict won't be present. Based on it releasing the cluster + * or mgmt_v3 lock */ + if (lock_ctx->dict == NULL) { + ret = glusterd_unlock (lock_ctx->uuid); + glusterd_op_unlock_send_resp (lock_ctx->req, ret); + } else { + ret = dict_get_str (lock_ctx->dict, "volname", &volname); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire volname"); + else { + ret = glusterd_mgmt_v3_unlock (volname, lock_ctx->uuid, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to release lock for %s", volname); + } - gf_log (this->name, GF_LOG_DEBUG, "Unlock Returned %d", ret); + glusterd_op_mgmt_v3_unlock_send_resp (lock_ctx->req, + &event->txn_id, ret); - glusterd_op_unlock_send_resp (lock_ctx->req, ret); + dict_unref (lock_ctx->dict); + } + + gf_log (this->name, GF_LOG_DEBUG, "Unlock Returned %d", ret); if (priv->pending_quorum_action) glusterd_do_quorum_action (); @@ -2399,7 +2848,8 @@ glusterd_op_ac_rcvd_lock_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, + &event->txn_id, NULL); gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -2603,9 +3053,6 @@ glusterd_op_build_payload (dict_t **req, char **op_errstr, dict_t *op_ctx) case GD_OP_STATEDUMP_VOLUME: case GD_OP_CLEARLOCKS_VOLUME: case GD_OP_DEFRAG_BRICK_VOLUME: -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: -#endif { ret = dict_get_str (dict, "volname", &volname); if (ret) { @@ -2834,7 +3281,8 @@ out: if (dict) dict_unref (dict); if (ret) { - glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL); + glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + &event->txn_id, NULL); opinfo.op_ret = ret; } @@ -2843,7 +3291,7 @@ out: opinfo.pending_count); if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_all_acc (); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); @@ -2852,10 +3300,10 @@ out: } static int32_t -glusterd_op_start_rb_timer (dict_t *dict) +glusterd_op_start_rb_timer (dict_t *dict, uuid_t *txn_id) { int32_t op = 0; - struct timeval timeout = {0, }; + struct timespec timeout = {0, }; glusterd_conf_t *priv = NULL; int32_t ret = -1; dict_t *rb_ctx = NULL; @@ -2871,12 +3319,12 @@ glusterd_op_start_rb_timer (dict_t *dict) } if (op != GF_REPLACE_OP_START) { - ret = glusterd_op_sm_inject_all_acc (); + ret = glusterd_op_sm_inject_all_acc (txn_id); goto out; } timeout.tv_sec = 5; - timeout.tv_usec = 0; + timeout.tv_nsec = 0; rb_ctx = dict_copy (dict, rb_ctx); @@ -2886,6 +3334,17 @@ glusterd_op_start_rb_timer (dict_t *dict) ret = -1; goto out; } + + ret = dict_set_bin (rb_ctx, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Failed to set transaction id."); + goto out; + } else + gf_log ("", GF_LOG_DEBUG, + "transaction_id = %s", uuid_utoa (*txn_id)); + priv->timer = gf_timer_call_after (THIS->ctx, timeout, glusterd_do_replace_brick, (void *) rb_ctx); @@ -2954,6 +3413,97 @@ out: return ret; } +static int +reassign_defrag_status (dict_t *dict, char *key, gf_defrag_status_t *status) +{ + int ret = 0; + + if (!*status) + return ret; + + switch (*status) { + case GF_DEFRAG_STATUS_STARTED: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED; + break; + + case GF_DEFRAG_STATUS_STOPPED: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_STOPPED; + break; + + case GF_DEFRAG_STATUS_COMPLETE: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE; + break; + + case GF_DEFRAG_STATUS_FAILED: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_FAILED; + break; + default: + break; + } + + ret = dict_set_int32(dict, key, *status); + if (ret) + gf_log (THIS->name, GF_LOG_WARNING, + "failed to reset defrag %s in dict", key); + + return ret; +} + +/* Check and reassign the defrag_status enum got from the rebalance process + * of all peers so that the rebalance-status CLI command can display if a + * full-rebalance or just a fix-layout was carried out. + */ +static int +glusterd_op_check_peer_defrag_status (dict_t *dict, int count) +{ + glusterd_volinfo_t *volinfo = NULL; + gf_defrag_status_t status = GF_DEFRAG_STATUS_NOT_STARTED; + char key[256] = {0,}; + char *volname = NULL; + int ret = -1; + int i = 1; + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, "Unable to get volume name"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, FMTSTR_CHECK_VOL_EXISTS, + volname); + goto out; + } + + if (volinfo->rebal.defrag_cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) { + /* Fix layout was not issued; we don't need to reassign + the status */ + ret = 0; + goto out; + } + + do { + memset (key, 0, 256); + snprintf (key, 256, "status-%d", i); + ret = dict_get_int32 (dict, key, (int32_t *)&status); + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, + "failed to get defrag %s", key); + goto out; + } + ret = reassign_defrag_status (dict, key, &status); + if (ret) + goto out; + i++; + } while (i <= count); + + ret = 0; +out: + return ret; + +} + /* This function is used to modify the op_ctx dict before sending it back * to cli. This is useful in situations like changing the peer uuids to * hostnames etc. @@ -3057,12 +3607,38 @@ glusterd_op_modify_op_ctx (glusterd_op_t op, void *ctx) goto out; } + /* add 'node-name-%d' into op_ctx with value uuid_str. + this will be used to convert to hostname later */ + { + char key[1024]; + char *uuid_str = NULL; + int i; + + for (i = 1; i <= count; i++) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "node-uuid-%d", i); + ret = dict_get_str (op_ctx, key, &uuid_str); + if (!ret) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), + "node-name-%d", i); + ret = dict_set_str (op_ctx, key, + uuid_str); + } + } + } + ret = glusterd_op_volume_dict_uuid_to_hostname (op_ctx, - "node-uuid-%d", + "node-name-%d", 1, (count + 1)); if (ret) gf_log (this->name, GF_LOG_WARNING, "Failed uuid to hostname conversion"); + + ret = glusterd_op_check_peer_defrag_status (op_ctx, count); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to reset defrag status for fix-layout"); break; default: @@ -3215,17 +3791,19 @@ out: if (dict) dict_unref (dict); if (ret) { - glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL); + glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + &event->txn_id, NULL); opinfo.op_ret = ret; } if (!opinfo.pending_count) { if (op == GD_OP_REPLACE_BRICK) { - ret = glusterd_op_start_rb_timer (op_dict); + ret = glusterd_op_start_rb_timer (op_dict, + &event->txn_id); } else { glusterd_op_modify_op_ctx (op, NULL); - ret = glusterd_op_sm_inject_all_acc (); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); } goto err; } @@ -3250,7 +3828,8 @@ glusterd_op_ac_rcvd_stage_op_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_ACC, + &event->txn_id, NULL); out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -3271,7 +3850,8 @@ glusterd_op_ac_stage_op_failed (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, NULL); out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -3292,7 +3872,8 @@ glusterd_op_ac_commit_op_failed (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, NULL); out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -3335,7 +3916,8 @@ glusterd_op_ac_brick_op_failed (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.brick_pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, ev_ctx->commit_ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, ev_ctx->commit_ctx); out: if (ev_ctx->rsp_dict) @@ -3377,7 +3959,7 @@ glusterd_op_ac_rcvd_commit_op_acc (glusterd_op_sm_event_t *event, void *ctx) goto out; } - ret = glusterd_op_start_rb_timer (op_ctx); + ret = glusterd_op_start_rb_timer (op_ctx, &event->txn_id); if (ret) { gf_log (this->name, GF_LOG_ERROR, "Couldn't start " "replace-brick operation."); @@ -3392,10 +3974,12 @@ glusterd_op_ac_rcvd_commit_op_acc (glusterd_op_sm_event_t *event, void *ctx) out: if (commit_ack_inject) { if (ret) - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + &event->txn_id, NULL); else if (!opinfo.pending_count) { glusterd_op_modify_op_ctx (op, NULL); - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC, + &event->txn_id, NULL); } /*else do nothing*/ } @@ -3416,7 +4000,8 @@ glusterd_op_ac_rcvd_unlock_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, + &event->txn_id, NULL); gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -3450,7 +4035,7 @@ glusterd_op_reset_ctx () } int32_t -glusterd_op_txn_complete () +glusterd_op_txn_complete (uuid_t *txn_id) { int32_t ret = -1; glusterd_conf_t *priv = NULL; @@ -3460,6 +4045,7 @@ glusterd_op_txn_complete () rpcsvc_request_t *req = NULL; void *ctx = NULL; char *op_errstr = NULL; + char *volname = NULL; xlator_t *this = NULL; this = THIS; @@ -3482,14 +4068,29 @@ glusterd_op_txn_complete () glusterd_op_reset_ctx (); glusterd_op_clear_errstr (); - ret = glusterd_unlock (MY_UUID); - - /* unlock cant/shouldnt fail here!! */ - if (ret) { - gf_log (this->name, GF_LOG_CRITICAL, - "Unable to clear local lock, ret: %d", ret); + /* Based on the op-version, we release the cluster or mgmt_v3 lock */ + if (priv->op_version < 3) { + ret = glusterd_unlock (MY_UUID); + /* unlock cant/shouldnt fail here!! */ + if (ret) + gf_log (this->name, GF_LOG_CRITICAL, + "Unable to clear local lock, ret: %d", ret); + else + gf_log (this->name, GF_LOG_DEBUG, "Cleared local lock"); } else { - gf_log (this->name, GF_LOG_DEBUG, "Cleared local lock"); + ret = dict_get_str (ctx, "volname", &volname); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire volname"); + + if (volname) { + ret = glusterd_mgmt_v3_unlock (volname, MY_UUID, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to release lock for %s", + volname); + } } ret = glusterd_op_send_cli_response (op, op_ret, @@ -3508,6 +4109,13 @@ glusterd_op_txn_complete () if (priv->pending_quorum_action) glusterd_do_quorum_action (); + + /* Clearing the transaction opinfo */ + ret = glusterd_clear_txn_opinfo (txn_id); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to clear transaction's opinfo"); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -3519,7 +4127,7 @@ glusterd_op_ac_unlocked_all (glusterd_op_sm_event_t *event, void *ctx) GF_ASSERT (event); - ret = glusterd_op_txn_complete (); + ret = glusterd_op_txn_complete (&event->txn_id); gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -3536,6 +4144,7 @@ glusterd_op_ac_stage_op (glusterd_op_sm_event_t *event, void *ctx) char *op_errstr = NULL; dict_t *dict = NULL; xlator_t *this = NULL; + uuid_t *txn_id = NULL; this = THIS; GF_ASSERT (this); @@ -3561,6 +4170,19 @@ glusterd_op_ac_stage_op (glusterd_op_sm_event_t *event, void *ctx) status); } + txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + + if (txn_id) + uuid_copy (*txn_id, event->txn_id); + else + gf_log (this->name, GF_LOG_ERROR, "Out of Memory"); + + ret = dict_set_bin (rsp_dict, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to set transaction id."); + ret = glusterd_op_stage_send_resp (req_ctx->req, req_ctx->op, status, op_errstr, rsp_dict); @@ -3625,6 +4247,7 @@ glusterd_op_ac_commit_op (glusterd_op_sm_event_t *event, void *ctx) dict_t *dict = NULL; dict_t *rsp_dict = NULL; xlator_t *this = NULL; + uuid_t *txn_id = NULL; this = THIS; GF_ASSERT (this); @@ -3654,10 +4277,22 @@ glusterd_op_ac_commit_op (glusterd_op_sm_event_t *event, void *ctx) "'Volume %s' failed: %d", gd_op_list[req_ctx->op], status); + txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + + if (txn_id) + uuid_copy (*txn_id, event->txn_id); + else + gf_log (this->name, GF_LOG_ERROR, "Out of Memory"); + + ret = dict_set_bin (rsp_dict, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to set transaction id."); + ret = glusterd_op_commit_send_resp (req_ctx->req, req_ctx->op, status, op_errstr, rsp_dict); - glusterd_op_fini_ctx (); if (op_errstr && (strcmp (op_errstr, ""))) GF_FREE (op_errstr); @@ -3686,7 +4321,6 @@ glusterd_op_ac_send_commit_failed (glusterd_op_sm_event_t *event, void *ctx) opinfo.op_ret, opinfo.op_errstr, op_ctx); - glusterd_op_fini_ctx (); if (opinfo.op_errstr && (strcmp (opinfo.op_errstr, ""))) { GF_FREE (opinfo.op_errstr); opinfo.op_errstr = NULL; @@ -3808,11 +4442,6 @@ glusterd_op_stage_validate (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_op_stage_clearlocks_volume (dict, op_errstr); break; -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: - ret = glusterd_op_stage_bd (dict, op_errstr); - break; -#endif case GD_OP_COPY_FILE: ret = glusterd_op_stage_copy_file (dict, op_errstr); @@ -3924,11 +4553,6 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_op_clearlocks_volume (dict, op_errstr, rsp_dict); break; -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: - ret = 0; - break; -#endif case GD_OP_COPY_FILE: ret = glusterd_op_copy_file (dict, op_errstr); @@ -3951,6 +4575,7 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, return ret; } + static int glusterd_bricks_select_stop_volume (dict_t *dict, char **op_errstr, struct list_head *selected) @@ -4267,24 +4892,95 @@ out: } int +get_replica_index_for_per_replica_cmd (glusterd_volinfo_t *volinfo, + dict_t *dict) { + int ret = 0; + char *hostname = NULL; + char *path = NULL; + int index = 0; + glusterd_brickinfo_t *brickinfo = NULL; + int cmd_replica_index = -1; + int replica_count = -1; + + + if (!dict) { + ret = -1; + goto out; + } + + ret = dict_get_str (dict, "per-replica-cmd-hostname", &hostname); + if (ret) + goto out; + ret = dict_get_str (dict, "per-replica-cmd-path", &path); + if (ret) + goto out; + + replica_count = volinfo->replica_count; + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (uuid_is_null (brickinfo->uuid)) + (void)glusterd_resolve_brick (brickinfo); + if (!strcmp (brickinfo->path, path) && + !strcmp (brickinfo->hostname, hostname)) { + cmd_replica_index = index/(replica_count); + goto out; + } + index++; + } + + +out: + if (ret) + cmd_replica_index = -1; + + return cmd_replica_index; +} + +int _select_rxlators_with_local_bricks (xlator_t *this, glusterd_volinfo_t *volinfo, - dict_t *dict) + dict_t *dict, cli_cmd_type type) { glusterd_brickinfo_t *brickinfo = NULL; glusterd_conf_t *priv = NULL; - int index = 1; + int index = 0; int rxlator_count = 0; int replica_count = 0; gf_boolean_t add = _gf_false; + int ret = 0; + int cmd_replica_index = -1; priv = this->private; replica_count = volinfo->replica_count; + + if (type == PER_REPLICA) { + + cmd_replica_index = get_replica_index_for_per_replica_cmd + (volinfo, dict); + if (cmd_replica_index == -1) { + ret = -1; + goto err; + } + } + + index = 1; + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { if (uuid_is_null (brickinfo->uuid)) (void)glusterd_resolve_brick (brickinfo); - if (!uuid_compare (MY_UUID, brickinfo->uuid)) - add = _gf_true; + switch (type) { + case ALL_REPLICA: + if (!uuid_compare (MY_UUID, brickinfo->uuid)) + add = _gf_true; + break; + case PER_REPLICA: + if (!uuid_compare (MY_UUID, brickinfo->uuid) && + ((index-1)/replica_count == cmd_replica_index)) + + add = _gf_true; + break; + } + if (index % replica_count == 0) { if (add) { _add_rxlator_to_dict (dict, volinfo->volname, @@ -4297,6 +4993,10 @@ _select_rxlators_with_local_bricks (xlator_t *this, glusterd_volinfo_t *volinfo, index++; } +err: + if (ret) + rxlator_count = -1; + return rxlator_count; } @@ -4337,9 +5037,10 @@ _select_rxlators_for_full_self_heal (xlator_t *this, return rxlator_count; } -#ifdef HAVE_BD_XLATOR + static int -glusterd_bricks_select_bd (dict_t *dict, char **op_errstr) +glusterd_bricks_select_snap (dict_t *dict, char **op_errstr, + struct list_head *selected) { int ret = -1; glusterd_conf_t *priv = NULL; @@ -4357,31 +5058,31 @@ glusterd_bricks_select_bd (dict_t *dict, char **op_errstr) ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (this->name, GF_LOG_ERROR, "Unable to get volname"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get" + " volname"); goto out; } ret = glusterd_volinfo_find (volname, &volinfo); if (ret) goto out; - pending_node = GF_CALLOC (1, sizeof (*pending_node), - gf_gld_mt_pending_node_t); - if (!pending_node) { - ret = -1; - goto out; - } - list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { brick_index++; if (uuid_compare (brickinfo->uuid, MY_UUID) || !glusterd_is_brick_started (brickinfo)) { continue; } + pending_node = GF_CALLOC (1, sizeof (*pending_node), + gf_gld_mt_pending_node_t); + if (!pending_node) { + ret = -1; + goto out; + } pending_node->node = brickinfo; pending_node->type = GD_NODE_BRICK; pending_node->index = brick_index; list_add_tail (&pending_node->list, - &opinfo.pending_bricks); + selected); pending_node = NULL; } @@ -4391,10 +5092,10 @@ out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning ret %d", ret); return ret; } -#endif static int -fill_shd_status_for_local_bricks (dict_t *dict, glusterd_volinfo_t *volinfo) +fill_shd_status_for_local_bricks (dict_t *dict, glusterd_volinfo_t *volinfo, + cli_cmd_type type, dict_t *req_dict) { glusterd_brickinfo_t *brickinfo = NULL; char msg[1024] = {0,}; @@ -4403,10 +5104,22 @@ fill_shd_status_for_local_bricks (dict_t *dict, glusterd_volinfo_t *volinfo) int index = 0; int ret = 0; xlator_t *this = NULL; + int cmd_replica_index = -1; this = THIS; snprintf (msg, sizeof (msg), "self-heal-daemon is not running on"); + if (type == PER_REPLICA) { + cmd_replica_index = get_replica_index_for_per_replica_cmd + (volinfo, req_dict); + if (cmd_replica_index == -1) { + gf_log (THIS->name, GF_LOG_ERROR, "Could not find the " + "replica index for per replica type command"); + ret = -1; + goto out; + } + } + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { if (uuid_is_null (brickinfo->uuid)) (void)glusterd_resolve_brick (brickinfo); @@ -4415,6 +5128,14 @@ fill_shd_status_for_local_bricks (dict_t *dict, glusterd_volinfo_t *volinfo) index++; continue; } + + if (type == PER_REPLICA) { + if (cmd_replica_index != (index/volinfo->replica_count)) { + index++; + continue; + } + + } snprintf (key, sizeof (key), "%d-status",index); snprintf (value, sizeof (value), "%s %s",msg, uuid_utoa(MY_UUID)); @@ -4483,21 +5204,49 @@ glusterd_bricks_select_heal_volume (dict_t *dict, char **op_errstr, goto out; } + switch (heal_op) { + case GF_AFR_OP_INDEX_SUMMARY: + case GF_AFR_OP_STATISTICS_HEAL_COUNT: + if (!glusterd_is_nodesvc_online ("glustershd")) { + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, "Received " + "empty ctx."); + goto out; + } - if (!glusterd_is_nodesvc_online ("glustershd") && - (heal_op == GF_AFR_OP_INDEX_SUMMARY)) { - - if (!rsp_dict) { - gf_log (this->name, GF_LOG_ERROR, "Received empty " - "ctx."); + ret = fill_shd_status_for_local_bricks (rsp_dict, + volinfo, + ALL_REPLICA, + dict); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "fill the shd status for the local " + "bricks"); goto out; + } + break; + case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA: + if (!glusterd_is_nodesvc_online ("glustershd")) { + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, "Received " + "empty ctx."); + goto out; + } + ret = fill_shd_status_for_local_bricks (rsp_dict, + volinfo, + PER_REPLICA, + dict); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "fill the shd status for the local" + " bricks."); + goto out; - ret = fill_shd_status_for_local_bricks (rsp_dict, volinfo); - if (ret) - gf_log (this->name, GF_LOG_ERROR, "Unable to fill the shd" - " status for the local bricks"); - goto out; + } + break; + default: + break; } @@ -4507,14 +5256,28 @@ glusterd_bricks_select_heal_volume (dict_t *dict, char **op_errstr, volinfo, dict); break; + case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA: + rxlator_count = _select_rxlators_with_local_bricks (this, + volinfo, + dict, + PER_REPLICA); + break; default: rxlator_count = _select_rxlators_with_local_bricks (this, volinfo, - dict); + dict, + ALL_REPLICA); break; } if (!rxlator_count) goto out; + if (rxlator_count == -1){ + gf_log (this->name, GF_LOG_ERROR, "Could not determine the" + "translator count"); + ret = -1; + goto out; + } + ret = dict_set_int32 (dict, "count", rxlator_count); if (ret) goto out; @@ -4776,7 +5539,8 @@ glusterd_op_ac_send_brick_op (glusterd_op_sm_event_t *event, void *ctx) if (!opinfo.pending_count && !opinfo.brick_pending_count) { glusterd_clear_pending_nodes (&opinfo.pending_bricks); - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, req_ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, req_ctx); } out: @@ -4830,7 +5594,8 @@ glusterd_op_ac_rcvd_brick_op_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.brick_pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, ev_ctx->commit_ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, &event->txn_id, + ev_ctx->commit_ctx); out: if (ev_ctx->rsp_dict) @@ -4882,11 +5647,9 @@ glusterd_op_bricks_select (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_bricks_select_rebalance_volume (dict, op_errstr, selected); break; -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: - ret = glusterd_bricks_select_bd (dict, op_errstr); + case GD_OP_SNAP: + ret = glusterd_bricks_select_snap (dict, op_errstr, selected); break; -#endif default: break; } @@ -5208,7 +5971,7 @@ glusterd_op_sm_new_event (glusterd_op_sm_event_type_t event_type, int glusterd_op_sm_inject_event (glusterd_op_sm_event_type_t event_type, - void *ctx) + uuid_t *txn_id, void *ctx) { int32_t ret = -1; glusterd_op_sm_event_t *event = NULL; @@ -5223,6 +5986,9 @@ glusterd_op_sm_inject_event (glusterd_op_sm_event_type_t event_type, event->ctx = ctx; + if (txn_id) + uuid_copy (event->txn_id, *txn_id); + gf_log (THIS->name, GF_LOG_DEBUG, "Enqueue event: '%s'", glusterd_op_sm_event_name_get (event->event)); list_add_tail (&event->list, &gd_op_sm_queue); @@ -5283,6 +6049,7 @@ glusterd_op_sm () glusterd_op_sm_t *state = NULL; glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; xlator_t *this = NULL; + glusterd_op_info_t txn_op_info; this = THIS; GF_ASSERT (this); @@ -5303,6 +6070,20 @@ glusterd_op_sm () "type: '%s'", glusterd_op_sm_event_name_get(event_type)); + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", + uuid_utoa (event->txn_id)); + + ret = glusterd_get_txn_opinfo (&event->txn_id, + &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get transaction's opinfo"); + glusterd_destroy_op_event_ctx (event); + GF_FREE (event); + continue; + } else + opinfo = txn_op_info; + state = glusterd_op_state_table[opinfo.state.state]; GF_ASSERT (state); @@ -5333,8 +6114,27 @@ glusterd_op_sm () return ret; } + if ((state[event_type].next_state == + GD_OP_STATE_DEFAULT) && + (event_type == GD_OP_EVENT_UNLOCK)) { + /* Clearing the transaction opinfo */ + ret = glusterd_clear_txn_opinfo(&event->txn_id); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to clear " + "transaction's opinfo"); + } else { + ret = glusterd_set_txn_opinfo (&event->txn_id, + &opinfo); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to set " + "transaction's opinfo"); + } + glusterd_destroy_op_event_ctx (event); GF_FREE (event); + } } @@ -5388,52 +6188,6 @@ glusterd_op_clear_op (glusterd_op_t op) } int32_t -glusterd_op_init_ctx (glusterd_op_t op) -{ - int ret = 0; - dict_t *dict = NULL; - xlator_t *this = NULL; - - this = THIS; - GF_ASSERT (this); - GF_ASSERT (GD_OP_NONE < op && op < GD_OP_MAX); - - if (_gf_false == glusterd_need_brick_op (op)) { - gf_log (this->name, GF_LOG_DEBUG, "Received op: %s, returning", - gd_op_list[op]); - goto out; - } - dict = dict_new (); - if (dict == NULL) { - ret = -1; - goto out; - } - ret = glusterd_op_set_ctx (dict); - if (ret) - goto out; -out: - gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); - return ret; -} - - - -int32_t -glusterd_op_fini_ctx () -{ - dict_t *dict = NULL; - - dict = glusterd_op_get_ctx (); - if (dict) - dict_unref (dict); - - glusterd_op_reset_ctx (); - return 0; -} - - - -int32_t glusterd_op_free_ctx (glusterd_op_t op, void *ctx) { @@ -5459,9 +6213,6 @@ glusterd_op_free_ctx (glusterd_op_t op, void *ctx) case GD_OP_STATEDUMP_VOLUME: case GD_OP_CLEARLOCKS_VOLUME: case GD_OP_DEFRAG_BRICK_VOLUME: -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: -#endif dict_unref (ctx); break; default: @@ -5490,4 +6241,3 @@ glusterd_op_sm_init () pthread_mutex_init (&gd_op_sm_lock, NULL); return 0; } - diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.h b/xlators/mgmt/glusterd/src/glusterd-op-sm.h index 321c5c484..cf57b78e0 100644 --- a/xlators/mgmt/glusterd/src/glusterd-op-sm.h +++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.h @@ -32,6 +32,8 @@ #include "protocol-common.h" #define GD_VOLUME_NAME_MAX 256 +#define GD_OP_PROTECTED (0x02) +#define GD_OP_UNPROTECTED (0x04) typedef enum glusterd_op_sm_state_ { GD_OP_STATE_DEFAULT = 0, @@ -75,6 +77,7 @@ struct glusterd_op_sm_event_ { struct list_head list; void *ctx; glusterd_op_sm_event_type_t event; + uuid_t txn_id; }; typedef struct glusterd_op_sm_event_ glusterd_op_sm_event_t; @@ -117,6 +120,7 @@ typedef struct glusterd_op_log_filename_ctx_ glusterd_op_log_filename_ctx_t; struct glusterd_op_lock_ctx_ { uuid_t uuid; + dict_t *dict; rpcsvc_request_t *req; }; @@ -162,12 +166,18 @@ typedef struct glusterd_gsync_status_temp { glusterd_volinfo_t *volinfo; char *node; }glusterd_gsync_status_temp_t; + +typedef enum cli_cmd_type_ { + PER_REPLICA, + ALL_REPLICA, + } cli_cmd_type; + int glusterd_op_sm_new_event (glusterd_op_sm_event_type_t event_type, glusterd_op_sm_event_t **new_event); int glusterd_op_sm_inject_event (glusterd_op_sm_event_type_t event_type, - void *ctx); + uuid_t *txn_id, void *ctx); int glusterd_op_sm_init (); @@ -251,10 +261,7 @@ glusterd_op_init_commit_rsp_dict (glusterd_op_t op); void glusterd_op_modify_op_ctx (glusterd_op_t op, void *op_ctx); -int32_t -glusterd_op_init_ctx (glusterd_op_t op); -int32_t -glusterd_op_fini_ctx (); + int32_t glusterd_volume_stats_read_perf (char *brick_path, int32_t blk_size, int32_t blk_count, double *throughput, double *time); @@ -278,4 +285,15 @@ glusterd_check_gsync_running (glusterd_volinfo_t *volinfo, gf_boolean_t *flag); int glusterd_defrag_volume_node_rsp (dict_t *req_dict, dict_t *rsp_dict, dict_t *op_ctx); +int +glusterd_is_valid_vg (glusterd_brickinfo_t *brick, int check_tag, char *msg); + +int32_t +glusterd_get_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t *opinfo); + +int32_t +glusterd_set_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t *opinfo); + +int32_t +glusterd_clear_txn_opinfo (uuid_t *txn_id); #endif diff --git a/xlators/mgmt/glusterd/src/glusterd-pmap.c b/xlators/mgmt/glusterd/src/glusterd-pmap.c index 9fdd782fe..a153ca1a9 100644 --- a/xlators/mgmt/glusterd/src/glusterd-pmap.c +++ b/xlators/mgmt/glusterd/src/glusterd-pmap.c @@ -1,5 +1,5 @@ /* - Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com> + Copyright (c) 2010-2013 Red Hat, Inc. <http://www.redhat.com> This file is part of GlusterFS. This file is licensed to you under your choice of the GNU Lesser @@ -52,8 +52,8 @@ pmap_port_isfree (int port) } -struct pmap_registry * -pmap_registry_new (void) +static struct pmap_registry * +pmap_registry_new (xlator_t *this) { struct pmap_registry *pmap = NULL; int i = 0; @@ -69,8 +69,8 @@ pmap_registry_new (void) pmap->ports[i].type = GF_PMAP_PORT_FOREIGN; } - pmap->base_port = GF_IANA_PRIV_PORTS_START; - pmap->last_alloc = GF_IANA_PRIV_PORTS_START; + pmap->base_port = pmap->last_alloc = + ((glusterd_conf_t *)(this->private))->base_port; return pmap; } @@ -86,7 +86,7 @@ pmap_registry_get (xlator_t *this) pmap = priv->pmap; if (!pmap) { - pmap = pmap_registry_new (); + pmap = pmap_registry_new (this); if (!pmap) return NULL; priv->pmap = pmap; diff --git a/xlators/mgmt/glusterd/src/glusterd-rebalance.c b/xlators/mgmt/glusterd/src/glusterd-rebalance.c index 60dd9c19b..b7b974c68 100644 --- a/xlators/mgmt/glusterd/src/glusterd-rebalance.c +++ b/xlators/mgmt/glusterd/src/glusterd-rebalance.c @@ -146,8 +146,6 @@ __glusterd_defrag_notify (struct rpc_clnt *rpc, void *mydata, GF_DEFRAG_STATUS_STARTED) { volinfo->rebal.defrag_status = GF_DEFRAG_STATUS_FAILED; - } else { - volinfo->rebal.defrag_cmd = 0; } } @@ -219,6 +217,7 @@ glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr, defrag->cmd = cmd; + volinfo->rebal.defrag_cmd = cmd; volinfo->rebal.op = op; LOCK_INIT (&defrag->lock); @@ -533,7 +532,7 @@ glusterd_op_stage_rebalance (dict_t *dict, char **op_errstr) case GF_DEFRAG_CMD_START: case GF_DEFRAG_CMD_START_LAYOUT_FIX: case GF_DEFRAG_CMD_START_FORCE: - if (is_origin_glusterd ()) { + if (is_origin_glusterd (dict)) { op_ctx = glusterd_op_get_ctx (); if (!op_ctx) { ret = -1; diff --git a/xlators/mgmt/glusterd/src/glusterd-replace-brick.c b/xlators/mgmt/glusterd/src/glusterd-replace-brick.c index 5c879b64c..54b830870 100644 --- a/xlators/mgmt/glusterd/src/glusterd-replace-brick.c +++ b/xlators/mgmt/glusterd/src/glusterd-replace-brick.c @@ -31,6 +31,7 @@ DEFAULT_VAR_RUN_DIRECTORY"/%s-"RB_CLIENT_MOUNTPOINT, \ volinfo->volname); +extern uuid_t global_txn_id; int glusterd_get_replace_op_str (gf1_cli_replace_op op, char *op_str) @@ -268,13 +269,6 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, goto out; } - if (volinfo->backend == GD_VOL_BK_BD) { - snprintf (msg, sizeof (msg), "replace brick not supported " - "for Block backend volume"); - *op_errstr = gf_strdup (msg); - goto out; - } - if (GLUSTERD_STATUS_STARTED != volinfo->status) { ret = -1; snprintf (msg, sizeof (msg), "volume: %s is not started", @@ -332,7 +326,7 @@ glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr, ret = -1; goto out; } - if (is_origin_glusterd ()) { + if (is_origin_glusterd (dict)) { if (!ctx) { ret = -1; gf_log (this->name, GF_LOG_ERROR, @@ -694,7 +688,7 @@ rb_src_brick_restart (glusterd_volinfo_t *volinfo, sleep (2); ret = glusterd_volume_start_glusterfs (volinfo, src_brickinfo, - _gf_false); + _gf_false); if (ret) { gf_log ("", GF_LOG_ERROR, "Unable to start " "glusterfs, ret: %d", ret); @@ -1638,7 +1632,7 @@ glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict) /* Set task-id, if available, in op_ctx dict for operations * other than start */ - if (is_origin_glusterd ()) { + if (is_origin_glusterd (dict)) { ctx = glusterd_op_get_ctx(); if (!ctx) { gf_log (this->name, GF_LOG_ERROR, "Failed to " @@ -1901,6 +1895,7 @@ glusterd_do_replace_brick (void *data) glusterd_brickinfo_t *src_brickinfo = NULL; glusterd_brickinfo_t *dst_brickinfo = NULL; glusterd_conf_t *priv = NULL; + uuid_t *txn_id = &global_txn_id; int ret = 0; @@ -1920,6 +1915,10 @@ glusterd_do_replace_brick (void *data) gf_log ("", GF_LOG_DEBUG, "Replace brick operation detected"); + ret = dict_get_bin (dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + ret = dict_get_int32 (dict, "operation", &op); if (ret) { gf_log ("", GF_LOG_DEBUG, @@ -2015,9 +2014,11 @@ glusterd_do_replace_brick (void *data) out: if (ret) - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + txn_id, NULL); else - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC, + txn_id, NULL); glusterd_op_sm (); } diff --git a/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c b/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c index 2d8d381bf..d5200a4ae 100644 --- a/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c +++ b/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c @@ -33,6 +33,7 @@ extern glusterd_op_info_t opinfo; +extern uuid_t global_txn_id; int32_t glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret, @@ -142,7 +143,7 @@ glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret, case GD_OP_LIST_VOLUME: case GD_OP_CLEARLOCKS_VOLUME: case GD_OP_HEAL_VOLUME: - case GD_OP_BD_OP: + case GD_OP_SNAP: { /*nothing specific to be done*/ break; @@ -581,6 +582,7 @@ __glusterd_cluster_lock_cbk (struct rpc_req *req, struct iovec *iov, glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; glusterd_peerinfo_t *peerinfo = NULL; xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; this = THIS; GF_ASSERT (this); @@ -625,7 +627,7 @@ out: event_type = GD_OP_EVENT_RCVD_ACC; } - ret = glusterd_op_sm_inject_event (event_type, NULL); + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); if (!ret) { glusterd_friend_sm (); @@ -644,6 +646,164 @@ glusterd_cluster_lock_cbk (struct rpc_req *req, struct iovec *iov, __glusterd_cluster_lock_cbk); } +static int32_t +glusterd_mgmt_v3_lock_peers_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + gd1_mgmt_v3_lock_rsp rsp = {{0},}; + int ret = -1; + int32_t op_ret = -1; + glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + if (-1 == req->rpc_status) { + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to decode mgmt_v3 lock " + "response received from peer"); + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; + } + +out: + op_ret = rsp.op_ret; + + txn_id = &rsp.txn_id; + + gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG, + "Received mgmt_v3 lock %s from uuid: %s", + (op_ret) ? "RJT" : "ACC", uuid_utoa (rsp.uuid)); + + ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, + "mgmt_v3 lock response received " + "from unknown peer: %s", uuid_utoa (rsp.uuid)); + } + + if (op_ret) { + event_type = GD_OP_EVENT_RCVD_RJT; + opinfo.op_ret = op_ret; + opinfo.op_errstr = gf_strdup ("Another transaction could be in " + "progress. Please try again after" + " sometime."); + } else { + event_type = GD_OP_EVENT_RCVD_ACC; + } + + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); + + if (!ret) { + glusterd_friend_sm (); + glusterd_op_sm (); + } + + GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe)); + return ret; +} + +int32_t +glusterd_mgmt_v3_lock_peers_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + glusterd_mgmt_v3_lock_peers_cbk_fn); +} + +static int32_t +glusterd_mgmt_v3_unlock_peers_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + gd1_mgmt_v3_unlock_rsp rsp = {{0},}; + int ret = -1; + int32_t op_ret = -1; + glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; + glusterd_peerinfo_t *peerinfo = NULL; + xlator_t *this = NULL; + uuid_t *txn_id = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + + if (-1 == req->rpc_status) { + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to decode mgmt_v3 unlock " + "response received from peer"); + rsp.op_ret = -1; + rsp.op_errno = EINVAL; + goto out; + } + +out: + op_ret = rsp.op_ret; + + txn_id = &rsp.txn_id; + + gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG, + "Received mgmt_v3 unlock %s from uuid: %s", + (op_ret) ? "RJT" : "ACC", + uuid_utoa (rsp.uuid)); + + ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); + + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, + "mgmt_v3 unlock response received " + "from unknown peer: %s", uuid_utoa (rsp.uuid)); + } + + if (op_ret) { + event_type = GD_OP_EVENT_RCVD_RJT; + opinfo.op_ret = op_ret; + opinfo.op_errstr = gf_strdup ("Another transaction could be in " + "progress. Please try again after" + " sometime."); + } else { + event_type = GD_OP_EVENT_RCVD_ACC; + } + + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); + + if (!ret) { + glusterd_friend_sm (); + glusterd_op_sm (); + } + + GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe)); + return ret; +} + +int32_t +glusterd_mgmt_v3_unlock_peers_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + glusterd_mgmt_v3_unlock_peers_cbk_fn); +} + int32_t __glusterd_cluster_unlock_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) @@ -654,6 +814,7 @@ __glusterd_cluster_unlock_cbk (struct rpc_req *req, struct iovec *iov, glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; glusterd_peerinfo_t *peerinfo = NULL; xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; this = THIS; GF_ASSERT (this); @@ -695,7 +856,7 @@ out: event_type = GD_OP_EVENT_RCVD_ACC; } - ret = glusterd_op_sm_inject_event (event_type, NULL); + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); if (!ret) { glusterd_friend_sm (); @@ -727,6 +888,7 @@ __glusterd_stage_op_cbk (struct rpc_req *req, struct iovec *iov, char err_str[2048] = {0}; char *peer_str = NULL; xlator_t *this = NULL; + uuid_t *txn_id = NULL; this = THIS; GF_ASSERT (this); @@ -779,6 +941,10 @@ out: "Received stage %s from uuid: %s", (op_ret) ? "RJT" : "ACC", uuid_utoa (rsp.uuid)); + ret = dict_get_bin (dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); if (ret) { @@ -814,7 +980,7 @@ out: break; } - ret = glusterd_op_sm_inject_event (event_type, NULL); + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); if (!ret) { glusterd_friend_sm (); @@ -854,6 +1020,8 @@ __glusterd_commit_op_cbk (struct rpc_req *req, struct iovec *iov, char err_str[2048] = {0}; char *peer_str = NULL; xlator_t *this = NULL; + uuid_t *txn_id = NULL; + this = THIS; GF_ASSERT (this); @@ -907,6 +1075,10 @@ __glusterd_commit_op_cbk (struct rpc_req *req, struct iovec *iov, "Received commit %s from uuid: %s", (op_ret)?"RJT":"ACC", uuid_utoa (rsp.uuid)); + ret = dict_get_bin (dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo); if (ret) { @@ -986,7 +1158,7 @@ __glusterd_commit_op_cbk (struct rpc_req *req, struct iovec *iov, } out: - ret = glusterd_op_sm_inject_event (event_type, NULL); + ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL); if (!ret) { glusterd_friend_sm (); @@ -1223,6 +1395,134 @@ out: } int32_t +glusterd_mgmt_v3_lock_peers (call_frame_t *frame, xlator_t *this, + void *data) +{ + gd1_mgmt_v3_lock_req req = {{0},}; + int ret = -1; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_conf_t *priv = NULL; + call_frame_t *dummy_frame = NULL; + dict_t *dict = NULL; + uuid_t *txn_id = NULL; + + if (!this) + goto out; + + dict = data; + + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_ptr (dict, "peerinfo", VOID (&peerinfo)); + if (ret) + goto out; + + //peerinfo should not be in payload + dict_del (dict, "peerinfo"); + + glusterd_get_uuid (&req.uuid); + + ret = dict_allocate_and_serialize (dict, &req.dict.dict_val, + &req.dict.dict_len); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to serialize dict " + "to request buffer"); + goto out; + } + + /* Sending valid transaction ID to peers */ + ret = dict_get_bin (dict, "transaction_id", + (void **)&txn_id); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get transaction id."); + goto out; + } else { + gf_log (this->name, GF_LOG_DEBUG, + "Transaction_id = %s", uuid_utoa (*txn_id)); + uuid_copy (req.txn_id, *txn_id); + } + + dummy_frame = create_frame (this, this->ctx->pool); + if (!dummy_frame) + goto out; + + ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame, + peerinfo->mgmt_v3, + GLUSTERD_MGMT_V3_LOCK, NULL, + this, glusterd_mgmt_v3_lock_peers_cbk, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_req); +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_mgmt_v3_unlock_peers (call_frame_t *frame, xlator_t *this, + void *data) +{ + gd1_mgmt_v3_unlock_req req = {{0},}; + int ret = -1; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_conf_t *priv = NULL; + call_frame_t *dummy_frame = NULL; + dict_t *dict = NULL; + uuid_t *txn_id = NULL; + + if (!this) + goto out; + + dict = data; + + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_ptr (dict, "peerinfo", VOID (&peerinfo)); + if (ret) + goto out; + + //peerinfo should not be in payload + dict_del (dict, "peerinfo"); + + glusterd_get_uuid (&req.uuid); + + ret = dict_allocate_and_serialize (dict, &req.dict.dict_val, + &req.dict.dict_len); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to serialize dict " + "to request buffer"); + goto out; + } + + /* Sending valid transaction ID to peers */ + ret = dict_get_bin (dict, "transaction_id", + (void **)&txn_id); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get transaction id."); + goto out; + } else { + gf_log (this->name, GF_LOG_DEBUG, + "Transaction_id = %s", uuid_utoa (*txn_id)); + uuid_copy (req.txn_id, *txn_id); + } + + dummy_frame = create_frame (this, this->ctx->pool); + if (!dummy_frame) + goto out; + + ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame, + peerinfo->mgmt_v3, + GLUSTERD_MGMT_V3_UNLOCK, NULL, + this, glusterd_mgmt_v3_unlock_peers_cbk, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_req); +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t glusterd_cluster_unlock (call_frame_t *frame, xlator_t *this, void *data) { @@ -1385,6 +1685,7 @@ __glusterd_brick_op_cbk (struct rpc_req *req, struct iovec *iov, glusterd_req_ctx_t *req_ctx = NULL; glusterd_pending_node_t *node = NULL; xlator_t *this = NULL; + uuid_t *txn_id = &global_txn_id; this = THIS; GF_ASSERT (this); @@ -1447,6 +1748,11 @@ __glusterd_brick_op_cbk (struct rpc_req *req, struct iovec *iov, } } out: + + ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + ev_ctx = GF_CALLOC (1, sizeof (*ev_ctx), gf_gld_mt_brick_rsp_ctx_t); GF_ASSERT (ev_ctx); if (op_ret) { @@ -1459,7 +1765,7 @@ out: ev_ctx->pending_node = frame->cookie; ev_ctx->rsp_dict = dict; ev_ctx->commit_ctx = frame->local; - ret = glusterd_op_sm_inject_event (event_type, ev_ctx); + ret = glusterd_op_sm_inject_event (event_type, txn_id, ev_ctx); if (!ret) { glusterd_friend_sm (); glusterd_op_sm (); @@ -1484,6 +1790,7 @@ int32_t glusterd_brick_op (call_frame_t *frame, xlator_t *this, void *data) { + gd1_mgmt_brick_op_req *req = NULL; int ret = 0; glusterd_conf_t *priv = NULL; @@ -1494,6 +1801,7 @@ glusterd_brick_op (call_frame_t *frame, xlator_t *this, glusterd_req_ctx_t *req_ctx = NULL; struct rpc_clnt *rpc = NULL; dict_t *op_ctx = NULL; + uuid_t *txn_id = &global_txn_id; if (!this) { ret = -1; @@ -1516,6 +1824,11 @@ glusterd_brick_op (call_frame_t *frame, xlator_t *this, goto out; } + + ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id); + + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id)); + list_for_each_entry (pending_node, &opinfo.pending_bricks, list) { dummy_frame = create_frame (this, this->ctx->pool); if (!dummy_frame) @@ -1593,7 +1906,8 @@ glusterd_brick_op (call_frame_t *frame, xlator_t *this, out: if (ret) { - glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, data); + glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + txn_id, data); opinfo.op_ret = ret; } gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -1621,6 +1935,12 @@ struct rpc_clnt_procedure gd_mgmt_actors[GLUSTERD_MGMT_MAXVALUE] = { [GLUSTERD_MGMT_COMMIT_OP] = {"COMMIT_OP", glusterd_commit_op}, }; +struct rpc_clnt_procedure gd_mgmt_v3_actors[GLUSTERD_MGMT_V3_MAXVALUE] = { + [GLUSTERD_MGMT_V3_NULL] = {"NULL", NULL }, + [GLUSTERD_MGMT_V3_LOCK] = {"MGMT_V3_LOCK", glusterd_mgmt_v3_lock_peers}, + [GLUSTERD_MGMT_V3_UNLOCK] = {"MGMT_V3_UNLOCK", glusterd_mgmt_v3_unlock_peers}, +}; + struct rpc_clnt_program gd_mgmt_prog = { .progname = "glusterd mgmt", .prognum = GD_MGMT_PROGRAM, @@ -1645,4 +1965,10 @@ struct rpc_clnt_program gd_peer_prog = { .numproc = GLUSTERD_FRIEND_MAXVALUE, }; - +struct rpc_clnt_program gd_mgmt_v3_prog = { + .progname = "glusterd mgmt v3", + .prognum = GD_MGMT_V3_PROGRAM, + .progver = GD_MGMT_V3_VERSION, + .proctable = gd_mgmt_v3_actors, + .numproc = GLUSTERD_MGMT_V3_MAXVALUE, +}; diff --git a/xlators/mgmt/glusterd/src/glusterd-sm.h b/xlators/mgmt/glusterd/src/glusterd-sm.h index e29bb7277..b9bedbe69 100644 --- a/xlators/mgmt/glusterd/src/glusterd-sm.h +++ b/xlators/mgmt/glusterd/src/glusterd-sm.h @@ -97,6 +97,7 @@ struct glusterd_peerinfo_ { struct rpc_clnt *rpc; rpc_clnt_prog_t *mgmt; rpc_clnt_prog_t *peer; + rpc_clnt_prog_t *mgmt_v3; int connected; gf_store_handle_t *shandle; glusterd_sm_tr_log_t sm_log; diff --git a/xlators/mgmt/glusterd/src/glusterd-snapshot.c b/xlators/mgmt/glusterd/src/glusterd-snapshot.c new file mode 100644 index 000000000..9b811cd05 --- /dev/null +++ b/xlators/mgmt/glusterd/src/glusterd-snapshot.c @@ -0,0 +1,5590 @@ +/* + Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ +#ifndef _CONFIG_H +#define _CONFIG_H +#include "config.h" +#endif + +#include <inttypes.h> +#include <sys/types.h> +#include <unistd.h> +#include <sys/resource.h> +#include <sys/statvfs.h> +#include <sys/mount.h> + +#include "globals.h" +#include "compat.h" +#include "protocol-common.h" +#include "xlator.h" +#include "logging.h" +#include "timer.h" +#include "glusterd-mem-types.h" +#include "glusterd.h" +#include "glusterd-sm.h" +#include "glusterd-op-sm.h" +#include "glusterd-utils.h" +#include "glusterd-store.h" +#include "run.h" +#include "glusterd-volgen.h" +#include "glusterd-mgmt.h" +#include "glusterd-syncop.h" + +#include "syscall.h" +#include "cli1-xdr.h" +#include "xdr-generic.h" + +#ifdef GF_LINUX_HOST_OS +#include <mntent.h> +#endif + +char snap_mount_folder[PATH_MAX]; + +static int32_t +glusterd_find_missed_snap (dict_t *rsp_dict, glusterd_volinfo_t *vol, + char *snap_uuid, struct list_head *peers, + int32_t op); + +/* This function will restore a snapshot volumes + * + * @param dict dictionary containing snapshot restore request + * @param op_errstr In case of any failure error message will be returned + * in this variable + * @return Negative value on Failure and 0 in success + */ +int +glusterd_snapshot_restore (dict_t *dict, char **op_errstr, dict_t *rsp_dict) +{ + int ret = -1; + char *volname = NULL; + char *snapname = NULL; + xlator_t *this = NULL; + glusterd_volinfo_t *snap_volinfo = NULL; + glusterd_volinfo_t *volinfo = NULL; + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "snap name"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (NULL == snap) { + ret = gf_asprintf (op_errstr, "Snap (%s) not found", + snapname); + if (ret < 0) { + goto out; + } + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); + ret = -1; + goto out; + } + + /* TODO : As of now there is only volume in snapshot. + * Change this when multiple volume snapshot is introduced + */ + snap_volinfo = list_entry (snap->volumes.next, glusterd_volinfo_t, + vol_list); + + ret = glusterd_volinfo_find (snap_volinfo->parent_volname, &volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not get volinfo of " + "%s", snap_volinfo->parent_volname); + goto out; + } + + if (is_origin_glusterd (dict) == _gf_true) { + /* From origin glusterd check if * + * any peers with snap bricks is down */ + ret = glusterd_find_missed_snap (rsp_dict, snap_volinfo, + snap_volinfo->volname, + &priv->peers, + GF_SNAP_OPTION_TYPE_RESTORE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to find missed snap restores"); + goto out; + } + } + + ret = gd_restore_snap_volume (rsp_dict, volinfo, snap_volinfo); + if (ret) { + /* No need to update op_errstr because it is assumed + * that the called function will do that in case of + * failure. + */ + gf_log (this->name, GF_LOG_ERROR, "Failed to restore " + "snap for %s volume", volname); + goto out; + } + + ret = 0; + + /* TODO: Need to check if we need to delete the snap after the + * operation is successful or not. Also need to persist the state + * of restore operation in the store. + */ +out: + return ret; +} + +/* This function is called before actual restore is taken place. This function + * will validate whether the snapshot volumes are ready to be restored or not. + * + * @param dict dictionary containing snapshot restore request + * @param op_errstr In case of any failure error message will be returned + * in this variable + * @param rsp_dict response dictionary + * @return Negative value on Failure and 0 in success + */ +int +glusterd_snapshot_restore_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int ret = -1; + int32_t i = 0; + int32_t volcount = 0; + gf_boolean_t snap_restored = _gf_false; + char key[PATH_MAX] = {0, }; + char *volname = NULL; + char *snapname = NULL; + glusterd_volinfo_t *volinfo = NULL; + glusterd_snap_t *snap = NULL; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "snap name"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (NULL == snap) { + ret = gf_asprintf (op_errstr, "Snap (%s) not found", + snapname); + if (ret < 0) { + goto out; + } + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); + ret = -1; + goto out; + } + + snap_restored = snap->snap_restored; + + if (snap_restored) { + ret = gf_asprintf (op_errstr, "Snap (%s) is already " + "restored", snapname); + if (ret < 0) { + goto out; + } + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); + ret = -1; + goto out; + } + + ret = dict_set_str (rsp_dict, "snapname", snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "snap name"); + goto out; + } + + ret = dict_get_int32 (dict, "volcount", &volcount); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get volume count"); + goto out; + } + + /* Snapshot restore will only work if all the volumes, + that are part of the snapshot, are stopped. */ + for (i = 1; i <= volcount; ++i) { + snprintf (key, sizeof (key), "volname%d", i); + ret = dict_get_str (dict, key, &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "get volume name"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + ret = gf_asprintf (op_errstr, "Volume (%s) not found", + volname); + if (ret < 0) { + goto out; + } + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); + ret = -1; + goto out; + } + + if (glusterd_is_volume_started (volinfo)) { + ret = gf_asprintf (op_errstr, "Volume (%s) has been " + "started. Volume needs to be stopped before restoring " + "a snapshot.", volname); + if (ret < 0) { + goto out; + } + gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr); + ret = -1; + goto out; + } + } + + ret = 0; +out: + return ret; +} + +int +snap_max_hard_limits_validate (dict_t *dict, char *volname, + uint64_t value, char **op_errstr) +{ + char err_str[PATH_MAX] = ""; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = -1; + uint64_t max_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + + if (volname) { + ret = glusterd_volinfo_find (volname, &volinfo); + if (!ret) { + if (volinfo->is_snap_volume) { + ret = -1; + snprintf (err_str, PATH_MAX, + "%s is a snap volume. Configuring " + "snap-max-hard-limit for a snap " + "volume is prohibited.", volname); + goto out; + } + } + } + + if (value) { + /* Max limit for the system is GLUSTERD_SNAPS_MAX_HARD_LIMIT + * but max limit for a volume is conf->snap_max_hard_limit. + */ + if (volname) { + max_limit = conf->snap_max_hard_limit; + } else { + max_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT; + } + } + + if ((value < 0) || (value > max_limit)) { + ret = -1; + snprintf (err_str, PATH_MAX, "Invalid snap-max-hard-limit" + "%"PRIu64 ". Expected range 0 - %"PRIu64, + value, max_limit); + goto out; + } + + ret = 0; +out: + if (ret) { + *op_errstr = gf_strdup (err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + } + return ret; +} + +int +glusterd_snapshot_config_prevalidate (dict_t *dict, char **op_errstr) +{ + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + xlator_t *this = NULL; + int ret = -1; + int config_command = 0; + char err_str[PATH_MAX] = {0,}; + glusterd_conf_t *conf = NULL; + uint64_t value = 0; + uint64_t hard_limit = 0; + uint64_t soft_limit = 0; + gf_loglevel_t loglevel = GF_LOG_ERROR; + uint64_t max_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + + ret = dict_get_int32 (dict, "config-command", &config_command); + if (ret) { + snprintf (err_str, sizeof (err_str), + "failed to get config-command type"); + goto out; + } + + ret = dict_get_uint64 (dict, "snap-max-hard-limit", &hard_limit); + + ret = dict_get_uint64 (dict, "snap-max-soft-limit", &soft_limit); + + ret = dict_get_str (dict, "volname", &volname); + + if (volname) { + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, sizeof (err_str), + "Volume %s does not exist.", volname); + goto out; + } + } + + switch (config_command) { + case GF_SNAP_CONFIG_TYPE_SET: + if (hard_limit) { + /* Validations for snap-max-hard-limits */ + ret = snap_max_hard_limits_validate (dict, volname, + hard_limit, op_errstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "snap-max-hard-limit validation " + "failed."); + goto out; + } + } + + if (soft_limit) { + max_limit = GLUSTERD_SNAPS_MAX_SOFT_LIMIT_PERCENT; + if ((soft_limit < 0) || (soft_limit > max_limit)) { + ret = -1; + snprintf (err_str, PATH_MAX, "Invalid " + "snap-max-soft-limit ""%" + PRIu64 ". Expected range 0 - %"PRIu64, + value, max_limit); + goto out; + } + break; + } + default: + break; + } + + ret = 0; +out: + + if (ret && err_str[0] != '\0') { + gf_log (this->name, loglevel, "%s", err_str); + *op_errstr = gf_strdup (err_str); + } + + return ret; +} + +int +glusterd_snap_create_pre_val_use_rsp_dict (dict_t *dst, dict_t *src) +{ + char *snap_brick_dir = NULL; + char *snap_device = NULL; + char *tmpstr = NULL; + char key[PATH_MAX] = ""; + char snapbrckcnt[PATH_MAX] = ""; + char snapbrckord[PATH_MAX] = ""; + int ret = -1; + int64_t i = -1; + int64_t j = -1; + int64_t volume_count = 0; + int64_t brick_count = 0; + int64_t brick_order = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dst); + GF_ASSERT (src); + + ret = dict_get_int64 (src, "volcount", &volume_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "get the volume count"); + goto out; + } + + for (i = 0; i < volume_count; i++) { + memset (snapbrckcnt, '\0', sizeof(snapbrckcnt)); + ret = snprintf (snapbrckcnt, sizeof(snapbrckcnt) - 1, + "vol%ld_brickcount", i+1); + ret = dict_get_int64 (src, snapbrckcnt, &brick_count); + if (ret) { + gf_log (this->name, GF_LOG_TRACE, + "No bricks for this volume in this dict"); + continue; + } + + for (j = 0; j < brick_count; j++) { + /* Fetching data from source dict */ + snprintf (key, sizeof(key) - 1, + "vol%ld.brickdir%ld", i+1, j); + + ret = dict_get_ptr (src, key, + (void **)&snap_brick_dir); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Unable to fetch %s", key); + continue; + } + + snprintf (key, sizeof(key) - 1, + "vol%ld.brick_snapdevice%ld", i+1, j); + + ret = dict_get_ptr (src, key, + (void **)&snap_device); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch snap_device"); + goto out; + } + + snprintf (snapbrckord, sizeof(snapbrckord) - 1, + "vol%ld.brick%ld.order", i+1, j); + + ret = dict_get_int64 (src, snapbrckord, &brick_order); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get brick order"); + goto out; + } + + /* Adding the data in the dst dict */ + snprintf (key, sizeof(key) - 1, + "vol%ld.brickdir%ld", i+1, brick_order); + + tmpstr = gf_strdup (snap_brick_dir); + if (!tmpstr) { + gf_log (this->name, GF_LOG_ERROR, + "Out Of Memory"); + ret = -1; + goto out; + } + ret = dict_set_dynstr (dst, key, tmpstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set %s", key); + GF_FREE (tmpstr); + goto out; + } + + snprintf (key, sizeof(key) - 1, + "vol%ld.brick_snapdevice%ld", + i+1, brick_order); + + tmpstr = gf_strdup (snap_device); + if (!tmpstr) { + ret = -1; + goto out; + } + ret = dict_set_dynstr (dst, key, tmpstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set %s", key); + GF_FREE (tmpstr); + goto out; + } + + } + } + + ret = 0; +out: + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_snap_pre_validate_use_rsp_dict (dict_t *dst, dict_t *src) +{ + int ret = -1; + int32_t snap_command = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + if (!dst || !src) { + gf_log (this->name, GF_LOG_ERROR, "Source or Destination " + "dict is empty."); + goto out; + } + + ret = dict_get_int32 (dst, "type", &snap_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case GF_SNAP_OPTION_TYPE_CREATE: + ret = glusterd_snap_create_pre_val_use_rsp_dict (dst, src); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to use " + "rsp dict"); + goto out; + } + break; + default: + break; + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int +glusterd_snapshot_create_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + char *volname = NULL; + char *snapname = NULL; + char *device = NULL; + char *tmpstr = NULL; + char *brick_dir = NULL; + char snap_brick_dir[PATH_MAX] = ""; + char *mnt_pt = NULL; + char key[PATH_MAX] = ""; + char snap_mount[PATH_MAX] = ""; + char snap_volname[64] = ""; + char err_str[PATH_MAX] = ""; + int ret = -1; + int64_t i = 0; + int64_t volcount = 0; + int64_t brick_count = 0; + int64_t brick_order = 0; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_volinfo_t *volinfo = NULL; + xlator_t *this = NULL; + uuid_t *snap_volid = NULL; + gf_loglevel_t loglevel = GF_LOG_ERROR; + glusterd_conf_t *conf = NULL; + int64_t effective_max_limit = 0; + + this = THIS; + GF_ASSERT (op_errstr); + conf = this->private; + GF_ASSERT (conf); + + ret = dict_get_int64 (dict, "volcount", &volcount); + if (ret) { + snprintf (err_str, sizeof (err_str), "Failed to " + "get the volume count"); + goto out; + } + if (volcount <= 0) { + snprintf (err_str, sizeof (err_str), "Invalid volume count %ld " + "supplied", volcount); + ret = -1; + goto out; + } + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + snprintf (err_str, sizeof (err_str), "Failed to get snapname"); + goto out; + } + + if (glusterd_find_snap_by_name (snapname)) { + ret = -1; + snprintf (err_str, sizeof (err_str), "Snap %s already exists", + snapname); + goto out; + } + + for (i = 1; i <= volcount; i++) { + snprintf (key, sizeof (key), "volname%ld", i); + ret = dict_get_str (dict, key, &volname); + if (ret) { + snprintf (err_str, sizeof (err_str), + "failed to get volume name"); + goto out; + } + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, sizeof (err_str), + "Volume (%s) does not exist ", volname); + goto out; + } + + ret = -1; + if (!glusterd_is_volume_started (volinfo)) { + snprintf (err_str, sizeof (err_str), "volume %s is " + "not started", volinfo->volname); + loglevel = GF_LOG_WARNING; + goto out; + } + if (glusterd_is_defrag_on (volinfo)) { + snprintf (err_str, sizeof (err_str), + "rebalance process is running for the " + "volume %s", volname); + loglevel = GF_LOG_WARNING; + goto out; + } + /* TODO: Also check whether geo replication is running */ + + if (volinfo->is_snap_volume == _gf_true) { + snprintf (err_str, sizeof (err_str), + "Volume %s is a snap volume", volname); + loglevel = GF_LOG_WARNING; + goto out; + } + + if (volinfo->snap_max_hard_limit < conf->snap_max_hard_limit) + effective_max_limit = volinfo->snap_max_hard_limit; + else + effective_max_limit = conf->snap_max_hard_limit; + + if (volinfo->snap_count >= effective_max_limit) { + snprintf (err_str, sizeof (err_str), + "The number of existing snaps has reached " + "the effective maximum limit of %"PRIu64" ," + "for the volume %s", effective_max_limit, + volname); + loglevel = GF_LOG_WARNING; + goto out; + } + + snprintf (key, sizeof(key) - 1, "vol%ld_volid", i); + ret = dict_get_bin (dict, key, (void **)&snap_volid); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch snap_volid"); + goto out; + } + + /* snap volume uuid is used as lvm snapshot name. + This will avoid restrictions on snapshot names + provided by user */ + GLUSTERD_GET_UUID_NOHYPHEN (snap_volname, *snap_volid); + + brick_count = 0; + brick_order = 0; + /* Adding snap bricks mount paths to the dict */ + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (uuid_compare (brickinfo->uuid, MY_UUID)) { + brick_order++; + continue; + } + + if (!glusterd_is_brick_started (brickinfo)) { + gf_log (this->name, GF_LOG_WARNING, + "brick %s:%s is not started", + brickinfo->hostname, + brickinfo->path); + brick_order++; + brick_count++; + continue; + } + + device = glusterd_get_brick_mount_details (brickinfo); + if (!device) { + snprintf (err_str, sizeof (err_str), + "getting device name for the brick " + "%s:%s failed", brickinfo->hostname, + brickinfo->path); + ret = -1; + goto out; + } + + device = glusterd_build_snap_device_path (device, + snap_volname); + if (!device) { + snprintf (err_str, sizeof (err_str), + "cannot copy the snapshot device " + "name (volname: %s, snapname: %s)", + volinfo->volname, snapname); + loglevel = GF_LOG_WARNING; + ret = -1; + goto out; + } + + snprintf (key, sizeof(key), + "vol%ld.brick_snapdevice%ld", i, + brick_count); + ret = dict_set_dynstr (rsp_dict, key, device); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set %s", key); + GF_FREE (device); + goto out; + } + + ret = glusterd_get_brick_root (brickinfo->path, + &mnt_pt); + if (ret) { + snprintf (err_str, sizeof (err_str), + "could not get the root of the brick path %s", + brickinfo->path); + loglevel = GF_LOG_WARNING; + goto out; + } + if (strncmp (brickinfo->path, mnt_pt, strlen(mnt_pt))) { + snprintf (err_str, sizeof (err_str), + "brick: %s brick mount: %s", + brickinfo->path, mnt_pt); + loglevel = GF_LOG_WARNING; + goto out; + } + + brick_dir = &brickinfo->path[strlen (mnt_pt)]; + brick_dir++; + + snprintf (snap_brick_dir, sizeof (snap_brick_dir), + "/%s", brick_dir); + + tmpstr = gf_strdup (snap_brick_dir); + if (!tmpstr) { + ret = -1; + goto out; + } + snprintf (key, sizeof(key), "vol%ld.brickdir%ld", i, + brick_count); + ret = dict_set_dynstr (rsp_dict, key, tmpstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set %s", snap_mount); + goto out; + } + tmpstr = NULL; + + snprintf (key, sizeof(key) - 1, "vol%ld.brick%ld.order", + i, brick_count); + ret = dict_set_int64 (rsp_dict, key, brick_order); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set %s", key); + goto out; + } + + brick_count++; + brick_order++; + } + snprintf (key, sizeof(key) - 1, "vol%ld_brickcount", i); + ret = dict_set_int64 (rsp_dict, key, brick_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set %s", + key); + goto out; + } + } + + ret = dict_set_int64 (rsp_dict, "volcount", volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set volcount"); + goto out; + } + + ret = 0; +out: + if (ret) + GF_FREE (tmpstr); + + if (ret && err_str[0] != '\0') { + gf_log (this->name, loglevel, "%s", err_str); + *op_errstr = gf_strdup (err_str); + } + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +glusterd_snap_t* +glusterd_new_snap_object() +{ + glusterd_snap_t *snap = NULL; + + snap = GF_CALLOC (1, sizeof (*snap), gf_gld_mt_snap_t); + + if (snap) { + if (LOCK_INIT (&snap->lock)) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed initiating" + " snap lock"); + GF_FREE (snap); + return NULL; + } + + INIT_LIST_HEAD (&snap->snap_list); + INIT_LIST_HEAD (&snap->volumes); + snap->snapname[0] = 0; + snap->snap_status = GD_SNAP_STATUS_INIT; + } + + return snap; + +}; + +/* Function glusterd_list_add_snapvol adds the volinfo object (snapshot volume) + to the snapshot object list and to the parent volume list */ +int32_t +glusterd_list_add_snapvol (glusterd_volinfo_t *origin_vol, + glusterd_volinfo_t *snap_vol) +{ + int ret = -1; + glusterd_snap_t *snap = NULL; + + GF_VALIDATE_OR_GOTO ("glusterd", origin_vol, out); + GF_VALIDATE_OR_GOTO ("glusterd", snap_vol, out); + + snap = snap_vol->snapshot; + GF_ASSERT (snap); + + list_add_tail (&snap_vol->vol_list, &snap->volumes); + LOCK (&origin_vol->lock); + { + list_add_order (&snap_vol->snapvol_list, + &origin_vol->snap_volumes, + glusterd_compare_snap_vol_time); + origin_vol->snap_count++; + } + UNLOCK (&origin_vol->lock); + + gf_log (THIS->name, GF_LOG_DEBUG, "Snap %s added to the list", + snap->snapname); + ret = 0; +out: + return ret; +} + +glusterd_snap_t* +glusterd_find_snap_by_name (char *snapname) +{ + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + GF_ASSERT (priv); + GF_ASSERT (snapname); + + + list_for_each_entry (snap, &priv->snapshots, snap_list) { + if (!strcmp (snap->snapname, snapname)) { + gf_log (THIS->name, GF_LOG_DEBUG, "Found " + "snap %s (%s)", snap->snapname, + uuid_utoa (snap->snap_id)); + goto out; + } + } + snap = NULL; +out: + return snap; +} + +glusterd_snap_t* +glusterd_find_snap_by_id (uuid_t snap_id) +{ + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + GF_ASSERT (priv); + + if (uuid_is_null(snap_id)) + goto out; + + list_for_each_entry (snap, &priv->snapshots, snap_list) { + if (!uuid_compare (snap->snap_id, snap_id)) { + gf_log (THIS->name, GF_LOG_DEBUG, "Found " + "snap %s (%s)", snap->snapname, + uuid_utoa (snap->snap_id)); + goto out; + } + } + snap = NULL; +out: + return snap; +} + +int +glusterd_do_lvm_snapshot_remove (glusterd_volinfo_t *snap_vol, + glusterd_brickinfo_t *brickinfo, + const char *mount_pt, const char *snap_device) +{ + int ret = -1; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + runner_t runner = {0,}; + char msg[1024] = {0, }; + char pidfile[PATH_MAX] = {0, }; + pid_t pid = -1; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + if (!brickinfo) { + gf_log (this->name, GF_LOG_ERROR, "brickinfo NULL"); + goto out; + } + + GF_ASSERT (snap_vol); + GF_ASSERT (mount_pt); + GF_ASSERT (snap_device); + + GLUSTERD_GET_BRICK_PIDFILE (pidfile, snap_vol, brickinfo, priv); + if (glusterd_is_service_running (pidfile, &pid)) { + ret = kill (pid, SIGKILL); + if (ret && errno != ESRCH) { + gf_log (this->name, GF_LOG_ERROR, "Unable to kill pid " + "%d reason : %s", pid, strerror(errno)); + goto out; + } + } + + runinit (&runner); + snprintf (msg, sizeof (msg), "umount the snapshot mounted path %s", + mount_pt); + runner_add_args (&runner, "umount", mount_pt, NULL); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + + /* We need not do synclock_unlock => runner_run => synclock_lock here. + Because it is needed if we are running a glusterfs process in + runner_run, so that when the glusterfs process started wants to + communicate to glusterd, glusterd wont be able to respond if it + has held the big lock. So we do unlock, run glusterfs process + (thus communicate to glusterd), lock. But since this is not a + glusterfs command that is being run, unlocking and then relocking + is not needed. + */ + ret = runner_run (&runner); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "unmounting the " + "path %s (brick: %s) failed (%s)", mount_pt, + brickinfo->path, strerror (errno)); + goto out; + } + + runinit (&runner); + snprintf (msg, sizeof(msg), "remove snapshot of the brick %s:%s, " + "device: %s", brickinfo->hostname, brickinfo->path, + snap_device); + runner_add_args (&runner, "/sbin/lvremove", "-f", snap_device, NULL); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + + ret = runner_run (&runner); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "removing snapshot of the " + "brick (%s:%s) of device %s failed", + brickinfo->hostname, brickinfo->path, snap_device); + goto out; + } + +out: + return ret; +} + +int32_t +glusterd_lvm_snapshot_remove (dict_t *rsp_dict, glusterd_volinfo_t *snap_vol) +{ + char *mnt_pt = NULL; + struct mntent *entry = NULL; + int32_t brick_count = -1; + int32_t ret = -1; + glusterd_brickinfo_t *brickinfo = NULL; + xlator_t *this = NULL; + FILE *mtab = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap_vol); + + if (!snap_vol) { + gf_log (this->name, GF_LOG_ERROR, "snap volinfo is NULL"); + goto out; + } + + brick_count = -1; + list_for_each_entry (brickinfo, &snap_vol->bricks, brick_list) { + brick_count++; + if (uuid_compare (brickinfo->uuid, MY_UUID)) + continue; + + if (brickinfo->snap_status == -1) { + gf_log (this->name, GF_LOG_INFO, + "snapshot was pending. lvm not present " + "for brick %s:%s of the snap %s.", + brickinfo->hostname, brickinfo->path, + snap_vol->snapshot->snapname); + + /* Adding missed delete to the dict */ + ret = glusterd_add_missed_snaps_to_dict + (rsp_dict, + snap_vol->volname, + brickinfo, + brick_count + 1, + GF_SNAP_OPTION_TYPE_DELETE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add missed snapshot info " + "for %s:%s in the rsp_dict", + brickinfo->hostname, + brickinfo->path); + goto out; + } + + continue; + } + + ret = glusterd_get_brick_root (brickinfo->path, &mnt_pt); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "getting the root " + "of the brick for volume %s (snap %s) failed ", + snap_vol->volname, snap_vol->snapshot->snapname); + goto out; + } + + entry = glusterd_get_mnt_entry_info (mnt_pt, mtab); + if (!entry) { + gf_log (this->name, GF_LOG_WARNING, "getting the mount" + " entry for the brick %s:%s of the snap %s " + "(volume: %s) failed", brickinfo->hostname, + brickinfo->path, snap_vol->snapshot->snapname, + snap_vol->volname); + ret = -1; + goto out; + } + ret = glusterd_do_lvm_snapshot_remove (snap_vol, brickinfo, + mnt_pt, + entry->mnt_fsname); + if (mtab) + endmntent (mtab); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "remove the snapshot %s (%s)", + brickinfo->path, entry->mnt_fsname); + goto out; + } + + } + + ret = 0; +out: + return ret; +} + +int32_t +glusterd_snap_volume_remove (dict_t *rsp_dict, + glusterd_volinfo_t *snap_vol, + gf_boolean_t remove_lvm, + gf_boolean_t force) +{ + int ret = -1; + int save_ret = 0; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_volinfo_t *origin_vol = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap_vol); + + if (!snap_vol) { + gf_log(this->name, GF_LOG_WARNING, "snap_vol in NULL"); + ret = -1; + goto out; + } + + list_for_each_entry (brickinfo, &snap_vol->bricks, brick_list) { + if (uuid_compare (brickinfo->uuid, MY_UUID)) + continue; + + ret = glusterd_brick_stop (snap_vol, brickinfo, _gf_false); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to stop " + "brick for volume %s", snap_vol->volname); + save_ret = ret; + + /* Continue to cleaning up the snap in case of error + if force flag is enabled */ + if (!force) + goto out; + } + } + + /* Only remove the backend lvm when required */ + if (remove_lvm) { + ret = glusterd_lvm_snapshot_remove (rsp_dict, snap_vol); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to remove " + "lvm snapshot volume %s", snap_vol->volname); + save_ret = ret; + if (!force) + goto out; + } + } + + ret = glusterd_store_delete_volume (snap_vol); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to remove volume %s " + "from store", snap_vol->volname); + save_ret = ret; + if (!force) + goto out; + } + + if (!list_empty(&snap_vol->snapvol_list)) { + ret = glusterd_volinfo_find (snap_vol->parent_volname, + &origin_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "parent volinfo %s for volume %s", + snap_vol->parent_volname, snap_vol->volname); + save_ret = ret; + if (!force) + goto out; + } + origin_vol->snap_count--; + } + + ret = glusterd_volinfo_delete (snap_vol); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to remove volinfo " + "%s ", snap_vol->volname); + save_ret = ret; + if (!force) + goto out; + } + + if (save_ret) + ret = save_ret; +out: + gf_log (this->name, GF_LOG_TRACE, "returning %d", ret); + return ret; +} + +int32_t +glusterd_snapobject_delete (glusterd_snap_t *snap) +{ + if (snap == NULL) { + gf_log(THIS->name, GF_LOG_WARNING, "snap is NULL"); + return -1; + } + + list_del_init (&snap->snap_list); + list_del_init (&snap->volumes); + if (LOCK_DESTROY(&snap->lock)) + gf_log (THIS->name, GF_LOG_WARNING, "Failed destroying lock" + "of snap %s", snap->snapname); + + GF_FREE (snap->description); + GF_FREE (snap); + + return 0; +} + +int32_t +glusterd_snap_remove (dict_t *rsp_dict, + glusterd_snap_t *snap, + gf_boolean_t remove_lvm, + gf_boolean_t force) +{ + int ret = -1; + int save_ret = 0; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_volinfo_t *tmp = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap); + + if (!snap) { + gf_log(this->name, GF_LOG_WARNING, "snap is NULL"); + ret = -1; + goto out; + } + + list_for_each_entry_safe (snap_vol, tmp, &snap->volumes, vol_list) { + ret = glusterd_snap_volume_remove (rsp_dict, snap_vol, + remove_lvm, force); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to remove " + "volinfo %s for snap %s", snap_vol->volname, + snap->snapname); + save_ret = ret; + + /* Continue to cleaning up the snap in case of error + if force flag is enabled */ + if (!force) + goto out; + } + } + + ret = glusterd_store_delete_snap (snap); + if (ret) { + gf_log(this->name, GF_LOG_WARNING, "Failed to remove snap %s " + "from store", snap->snapname); + save_ret = ret; + if (!force) + goto out; + } + + ret = glusterd_snapobject_delete (snap); + if (ret) + gf_log (this->name, GF_LOG_WARNING, "Failed to delete " + "snap object %s", snap->snapname); + + if (save_ret) + ret = save_ret; +out: + gf_log (THIS->name, GF_LOG_TRACE, "returning %d", ret); + return ret; +} + +static int +glusterd_snapshot_get_snapvol_detail (dict_t *dict, + glusterd_volinfo_t *snap_vol, + char *keyprefix, int detail) +{ + int ret = -1; + int snap_limit = 0; + char key[PATH_MAX] = {0,}; + char *value = NULL; + glusterd_volinfo_t *origin_vol = NULL; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + + this = THIS; + conf = this->private; + GF_ASSERT (conf); + + GF_ASSERT (dict); + GF_ASSERT (snap_vol); + GF_ASSERT (keyprefix); + + /* Volume Name */ + value = gf_strdup (snap_vol->volname); + if (!value) + goto out; + + snprintf (key, sizeof (key), "%s.volname", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "volume name in dictionary: %s", key); + goto out; + } + + /* Volume ID */ + value = gf_strdup (uuid_utoa (snap_vol->volume_id)); + if (NULL == value) { + ret = -1; + goto out; + } + + snprintf (key, sizeof (key), "%s.vol-id", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "volume id in dictionary: %s", key); + goto out; + } + value = NULL; + + /* volume status */ + snprintf (key, sizeof (key), "%s.vol-status", keyprefix); + switch (snap_vol->status) { + case GLUSTERD_STATUS_STARTED: + ret = dict_set_str (dict, key, "Started"); + break; + case GLUSTERD_STATUS_STOPPED: + ret = dict_set_str (dict, key, "Stopped"); + break; + case GD_SNAP_STATUS_NONE: + ret = dict_set_str (dict, key, "None"); + break; + default: + gf_log (this->name, GF_LOG_ERROR, "Invalid volume status"); + ret = -1; + goto out; + } + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set volume status" + " in dictionary: %s", key); + goto out; + } + + + ret = glusterd_volinfo_find (snap_vol->parent_volname, &origin_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to get the parent " + "volinfo for the volume %s", snap_vol->volname); + goto out; + } + + /* Snaps available */ + if (conf->snap_max_hard_limit < origin_vol->snap_max_hard_limit) { + snap_limit = conf->snap_max_hard_limit; + gf_log(this->name, GF_LOG_DEBUG, "system snap-max-hard-limit is" + " lesser than volume snap-max-hard-limit, " + "snap-max-hard-limit value is set to %d", snap_limit); + } else { + snap_limit = origin_vol->snap_max_hard_limit; + gf_log(this->name, GF_LOG_DEBUG, "volume snap-max-hard-limit is" + " lesser than system snap-max-hard-limit, " + "snap-max-hard-limit value is set to %d", snap_limit); + } + + snprintf (key, sizeof (key), "%s.snaps-available", keyprefix); + if (snap_limit > origin_vol->snap_count) + ret = dict_set_int32 (dict, key, + snap_limit - origin_vol->snap_count); + else + ret = dict_set_int32 (dict, key, 0); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set available snaps"); + goto out; + } + + snprintf (key, sizeof (key), "%s.snapcount", keyprefix); + ret = dict_set_int32 (dict, key, origin_vol->snap_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not save snapcount"); + goto out; + } + + if (!detail) + goto out; + + /* Parent volume name */ + value = gf_strdup (snap_vol->parent_volname); + if (!value) + goto out; + + snprintf (key, sizeof (key), "%s.origin-volname", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set parent " + "volume name in dictionary: %s", key); + goto out; + } + value = NULL; + + ret = 0; +out: + if (value) + GF_FREE (value); + + return ret; +} + +static int +glusterd_snapshot_get_snap_detail (dict_t *dict, glusterd_snap_t *snap, + char *keyprefix, glusterd_volinfo_t *volinfo) +{ + int ret = -1; + int volcount = 0; + char key[PATH_MAX] = {0,}; + char *value = NULL; + char *timestr = NULL; + struct tm *tmptr = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_volinfo_t *tmp_vol = NULL; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (dict); + GF_ASSERT (snap); + GF_ASSERT (keyprefix); + + /* Snap Name */ + value = gf_strdup (snap->snapname); + if (!value) + goto out; + + snprintf (key, sizeof (key), "%s.snapname", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "snap name in dictionary"); + goto out; + } + + /* Snap ID */ + value = gf_strdup (uuid_utoa (snap->snap_id)); + if (NULL == value) { + ret = -1; + goto out; + } + + snprintf (key, sizeof (key), "%s.snap-id", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "snap id in dictionary"); + goto out; + } + value = NULL; + + tmptr = localtime (&(snap->time_stamp)); + if (NULL == tmptr) { + gf_log (this->name, GF_LOG_ERROR, "Failed to convert " + "time_t to *tm"); + ret = -1; + goto out; + } + + timestr = GF_CALLOC (1, PATH_MAX, gf_gld_mt_char); + if (NULL == timestr) { + ret = -1; + goto out; + } + + ret = strftime (timestr, PATH_MAX, "%Y-%m-%d %H:%M:%S", tmptr); + if (0 == ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to convert time_t " + "to string"); + ret = -1; + goto out; + } + + snprintf (key, sizeof (key), "%s.snap-time", keyprefix); + ret = dict_set_dynstr (dict, key, timestr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "snap time stamp in dictionary"); + goto out; + } + timestr = NULL; + + /* If snap description is provided then add that into dictionary */ + if (NULL != snap->description) { + value = gf_strdup (snap->description); + if (NULL == value) { + ret = -1; + goto out; + } + + snprintf (key, sizeof (key), "%s.snap-desc", keyprefix); + ret = dict_set_dynstr (dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "snap description in dictionary"); + goto out; + } + value = NULL; + } + + snprintf (key, sizeof (key), "%s.snap-status", keyprefix); + switch (snap->snap_status) { + case GD_SNAP_STATUS_INIT: + ret = dict_set_str (dict, key, "Init"); + break; + case GD_SNAP_STATUS_IN_USE: + ret = dict_set_str (dict, key, "In-use"); + break; + case GD_SNAP_STATUS_DECOMMISSION: + ret = dict_set_str (dict, key, "Decommisioned"); + break; + case GD_SNAP_STATUS_RESTORED: + ret = dict_set_str (dict, key, "Restored"); + break; + case GD_SNAP_STATUS_NONE: + ret = dict_set_str (dict, key, "None"); + break; + default: + gf_log (this->name, GF_LOG_ERROR, "Invalid snap status"); + ret = -1; + goto out; + } + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snap status " + "in dictionary"); + goto out; + } + + if (volinfo) { + volcount = 1; + snprintf (key, sizeof (key), "%s.vol%d", keyprefix, volcount); + ret = glusterd_snapshot_get_snapvol_detail (dict, + volinfo, key, 0); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "get volume detail %s for snap %s", + snap_vol->volname, snap->snapname); + goto out; + } + goto done; + } + + list_for_each_entry_safe (snap_vol, tmp_vol, &snap->volumes, vol_list) { + volcount++; + snprintf (key, sizeof (key), "%s.vol%d", keyprefix, volcount); + ret = glusterd_snapshot_get_snapvol_detail (dict, + snap_vol, key, 1); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "get volume detail %s for snap %s", + snap_vol->volname, snap->snapname); + goto out; + } + } + +done: + snprintf (key, sizeof (key), "%s.vol-count", keyprefix); + ret = dict_set_int32 (dict, key, volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set %s", + key); + goto out; + } + + ret = 0; +out: + if (value) + GF_FREE (value); + + if (timestr) + GF_FREE(timestr); + + return ret; +} + +static int +glusterd_snapshot_get_all_snap_info (dict_t *dict) +{ + int ret = -1; + int snapcount = 0; + char key[PATH_MAX] = {0,}; + glusterd_snap_t *snap = NULL; + glusterd_snap_t *tmp_snap = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + + /* General parameter validation */ + GF_ASSERT (dict); + + list_for_each_entry_safe (snap, tmp_snap, &priv->snapshots, snap_list) { + snapcount++; + snprintf (key, sizeof (key), "snap%d", snapcount); + ret = glusterd_snapshot_get_snap_detail (dict, snap, key, NULL); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "snapdetail for snap %s", snap->snapname); + goto out; + } + } + + ret = dict_set_int32 (dict, "snap-count", snapcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount"); + goto out; + } + + ret = 0; +out: + return ret; +} + +int +glusterd_snapshot_get_info_by_volume (dict_t *dict, char *volname, + char *err_str, size_t len) +{ + int ret = -1; + int snapcount = 0; + int snap_limit = 0; + char *value = NULL; + char key[PATH_MAX] = ""; + glusterd_volinfo_t *volinfo = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_volinfo_t *tmp_vol = NULL; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + + this = THIS; + conf = this->private; + GF_ASSERT (conf); + + GF_ASSERT (dict); + GF_ASSERT (volname); + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, len, "Volume (%s) does not exist", volname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + goto out; + } + + /* Snaps available */ + if (conf->snap_max_hard_limit < volinfo->snap_max_hard_limit) { + snap_limit = conf->snap_max_hard_limit; + gf_log(this->name, GF_LOG_DEBUG, "system snap-max-hard-limit is" + " lesser than volume snap-max-hard-limit, " + "snap-max-hard-limit value is set to %d", snap_limit); + } else { + snap_limit = volinfo->snap_max_hard_limit; + gf_log(this->name, GF_LOG_DEBUG, "volume snap-max-hard-limit is" + " lesser than system snap-max-hard-limit, " + "snap-max-hard-limit value is set to %d", snap_limit); + } + + if (snap_limit > volinfo->snap_count) + ret = dict_set_int32 (dict, "snaps-available", + snap_limit - volinfo->snap_count); + else + ret = dict_set_int32 (dict, "snaps-available", 0); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set available snaps"); + goto out; + } + + /* Origin volume name */ + value = gf_strdup (volinfo->volname); + if (!value) + goto out; + + ret = dict_set_dynstr (dict, "origin-volname", value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set parent " + "volume name in dictionary: %s", key); + goto out; + } + value = NULL; + + list_for_each_entry_safe (snap_vol, tmp_vol, &volinfo->snap_volumes, + snapvol_list) { + snapcount++; + snprintf (key, sizeof (key), "snap%d", snapcount); + ret = glusterd_snapshot_get_snap_detail (dict, + snap_vol->snapshot, + key, snap_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "snapdetail for snap %s", + snap_vol->snapshot->snapname); + goto out; + } + } + ret = dict_set_int32 (dict, "snap-count", snapcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount"); + goto out; + } + + ret = 0; +out: + if (value) + GF_FREE (value); + + return ret; +} + +/* This function will be called from RPC handler routine. + * This function is responsible for getting the requested + * snapshot info into the dictionary. + * + * @param req RPC request object. Required for sending a response back. + * @param op glusterd operation. Required for sending a response back. + * @param dict pointer to dictionary which will contain both + * request and response key-pair values. + * @return -1 on error and 0 on success + */ +int +glusterd_handle_snapshot_info (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + int8_t snap_driven = 1; + char *volname = NULL; + char *snapname = NULL; + glusterd_snap_t *snap = NULL; + xlator_t *this = NULL; + int32_t cmd = GF_SNAP_INFO_TYPE_ALL; + + this = THIS; + GF_ASSERT (this); + + GF_VALIDATE_OR_GOTO (this->name, req, out); + GF_VALIDATE_OR_GOTO (this->name, dict, out); + + + ret = dict_get_int32 (dict, "cmd", &cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get type " + "of snapshot info"); + goto out; + } + + switch (cmd) { + case GF_SNAP_INFO_TYPE_ALL: + { + ret = glusterd_snapshot_get_all_snap_info (dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get info of all snaps"); + goto out; + } + break; + } + + case GF_SNAP_INFO_TYPE_SNAP: + { + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get snap name"); + goto out; + } + + ret = dict_set_int32 (dict, "snap-count", 1); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set snapcount"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + snprintf (err_str, len, + "Snap (%s) does not exist", snapname); + gf_log (this->name, GF_LOG_ERROR, + "%s", err_str); + ret = -1; + goto out; + } + ret = glusterd_snapshot_get_snap_detail (dict, snap, + "snap1", NULL); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get snap detail of snap " + "%s", snap->snapname); + goto out; + } + break; + } + + case GF_SNAP_INFO_TYPE_VOL: + { + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get volname"); + goto out; + } + ret = glusterd_snapshot_get_info_by_volume (dict, + volname, err_str, len); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get volume info of volume " + "%s", volname); + goto out; + } + snap_driven = 0; + break; + } + } + + ret = dict_set_int8 (dict, "snap-driven", snap_driven); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snap-driven"); + goto out; + } + + /* If everything is successful then send the response back to cli. + * In case of failure the caller of this function will take care + of the response */ + ret = glusterd_op_send_cli_response (op, 0, 0, req, dict, err_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to send cli " + "response"); + goto out; + } + + ret = 0; + +out: + return ret; +} + +/* This function sets all the snapshot names in the dictionary */ +int +glusterd_snapshot_get_all_snapnames (dict_t *dict) +{ + int ret = -1; + int snapcount = 0; + char *snapname = NULL; + char key[PATH_MAX] = {0,}; + glusterd_snap_t *snap = NULL; + glusterd_snap_t *tmp_snap = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (dict); + + list_for_each_entry_safe (snap, tmp_snap, &priv->snapshots, snap_list) { + snapcount++; + snapname = gf_strdup (snap->snapname); + if (!snapname) { + gf_log (this->name, GF_LOG_ERROR, "strdup failed"); + ret = -1; + goto out; + } + snprintf (key, sizeof (key), "snapname%d", snapcount); + ret = dict_set_dynstr (dict, key, snapname); + if (ret) { + GF_FREE (snapname); + gf_log (this->name, GF_LOG_ERROR, "Failed to set %s", + key); + goto out; + } + } + + ret = dict_set_int32 (dict, "snap-count", snapcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount"); + goto out; + } + + ret = 0; +out: + + return ret; +} + +/* This function sets all the snapshot names + under a given volume in the dictionary */ +int +glusterd_snapshot_get_vol_snapnames (dict_t *dict, glusterd_volinfo_t *volinfo) +{ + int ret = -1; + int snapcount = 0; + char *snapname = NULL; + char key[PATH_MAX] = {0,}; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_volinfo_t *tmp_vol = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (dict); + GF_ASSERT (volinfo); + + list_for_each_entry_safe (snap_vol, tmp_vol, + &volinfo->snap_volumes, snapvol_list) { + snapcount++; + snapname = gf_strdup (snap_vol->snapshot->snapname); + if (!snapname) { + gf_log (this->name, GF_LOG_ERROR, + "strdup failed"); + ret = -1; + goto out; + } + snprintf (key, sizeof (key), "snapname%d", snapcount); + ret = dict_set_dynstr (dict, key, snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "set %s", key); + GF_FREE (snapname); + goto out; + } + } + + ret = dict_set_int32 (dict, "snap-count", snapcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount"); + goto out; + } + + ret = 0; +out: + + return ret; +} + +int +glusterd_handle_snapshot_list (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + + GF_VALIDATE_OR_GOTO (this->name, req, out); + GF_VALIDATE_OR_GOTO (this->name, dict, out); + + /* Ignore error for getting volname as it is optional */ + ret = dict_get_str (dict, "volname", &volname); + + if (NULL == volname) { + ret = glusterd_snapshot_get_all_snapnames (dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get snapshot list"); + goto out; + } + } else { + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, len, + "Volume (%s) does not exist", volname); + gf_log (this->name, GF_LOG_ERROR, + "%s", err_str); + goto out; + } + + ret = glusterd_snapshot_get_vol_snapnames (dict, volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get snapshot list for volume %s", + volname); + goto out; + } + } + + /* If everything is successful then send the response back to cli. + In case of failure the caller of this function will take of response.*/ + ret = glusterd_op_send_cli_response (op, 0, 0, req, dict, err_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to send cli " + "response"); + goto out; + } + + ret = 0; + +out: + return ret; +} + +/* This is a snapshot create handler function. This function will be + * executed in the originator node. This function is responsible for + * calling mgmt_v3 framework to do the actual snap creation on all the bricks + * + * @param req RPC request object + * @param op gluster operation + * @param dict dictionary containing snapshot restore request + * @param err_str In case of an err this string should be populated + * @param len length of err_str buffer + * + * @return Negative value on Failure and 0 in success + */ +int +glusterd_handle_snapshot_create (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + char *volname = NULL; + char *snapname = NULL; + int64_t volcount = 0; + xlator_t *this = NULL; + char key[PATH_MAX] = ""; + char *username = NULL; + char *password = NULL; + uuid_t *uuid_ptr = NULL; + uuid_t tmp_uuid = {0}; + int i = 0; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (req); + GF_ASSERT (dict); + GF_ASSERT (err_str); + + ret = dict_get_int64 (dict, "volcount", &volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "get the volume count"); + goto out; + } + if (volcount <= 0) { + gf_log (this->name, GF_LOG_ERROR, "Invalid volume count %ld " + "supplied", volcount); + ret = -1; + goto out; + } + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to get the snapname"); + goto out; + } + + if (strlen(snapname) >= GLUSTERD_MAX_SNAP_NAME) { + snprintf (err_str, len, "snapname cannot exceed 255 " + "characters"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + ret = -1; + goto out; + } + + uuid_ptr = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + if (!uuid_ptr) { + gf_log (this->name, GF_LOG_ERROR, "Out Of Memory"); + ret = -1; + goto out; + } + + uuid_generate (*uuid_ptr); + ret = dict_set_bin (dict, "snap-id", uuid_ptr, sizeof(uuid_t)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to set snap-id"); + GF_FREE (uuid_ptr); + goto out; + } + uuid_ptr = NULL; + + ret = dict_set_int64 (dict, "snap-time", (int64_t)time(NULL)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to set snap-time"); + goto out; + } + + for (i = 1; i <= volcount; i++) { + snprintf (key, sizeof (key), "volname%d", i); + ret = dict_get_str (dict, key, &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get volume name"); + goto out; + } + + /* generate internal username and password for the snap*/ + uuid_generate (tmp_uuid); + username = gf_strdup (uuid_utoa (tmp_uuid)); + snprintf (key, sizeof(key), "volume%d_username", i); + ret = dict_set_dynstr (dict, key, username); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snap " + "username for volume %s", volname); + GF_FREE (username); + goto out; + } + + uuid_generate (tmp_uuid); + password = gf_strdup (uuid_utoa (tmp_uuid)); + snprintf (key, sizeof(key), "volume%d_password", i); + ret = dict_set_dynstr (dict, key, password); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set snap " + "password for volume %s", volname); + GF_FREE (password); + goto out; + } + + uuid_ptr = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + if (!uuid_ptr) { + gf_log (this->name, GF_LOG_ERROR, "Out Of Memory"); + ret = -1; + goto out; + } + + snprintf (key, sizeof(key) - 1, "vol%d_volid", i); + uuid_generate (*uuid_ptr); + ret = dict_set_bin (dict, key, uuid_ptr, sizeof(uuid_t)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set snap_volid"); + GF_FREE (uuid_ptr); + goto out; + } + } + + ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to initiate snap " + "phases"); + } + +out: + return ret; +} + +/* This is a snapshot status handler function. This function will be + * executed in a originator node. This function is responsible for + * calling mgmt v3 framework to get the actual snapshot status from + * all the bricks + * + * @param req RPC request object + * @param op gluster operation + * @param dict dictionary containing snapshot status request + * @param err_str In case of an err this string should be populated + * @param len length of err_str buffer + * + * return : 0 in case of success. + * -1 in case of failure. + * + */ +int +glusterd_handle_snapshot_status (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + char *volname = NULL; + char *snapname = NULL; + char *buf = NULL; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + int32_t cmd = -1; + int i = 0; + dict_t *voldict = NULL; + char key[PATH_MAX] = ""; + glusterd_volinfo_t *volinfo = NULL; + glusterd_snap_t *snap = NULL; + glusterd_volinfo_t *snap_volinfo = NULL; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + + GF_ASSERT (conf); + GF_ASSERT (req); + GF_ASSERT (dict); + GF_ASSERT (err_str); + + ret = dict_get_int32 (dict, "cmd", &cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not get status type"); + goto out; + } + switch (cmd) { + case GF_SNAP_STATUS_TYPE_ALL: + { + /* IF we give "gluster snapshot status" + * then lock is held on all snaps. + * This is the place where necessary information + * (snapname and snapcount)is populated in dictionary + * for locking. + */ + ++i; + list_for_each_entry (snap, &conf->snapshots, snap_list) + { + snprintf (key, sizeof (key), "snapname%d", i); + buf = gf_strdup (snap->snapname); + if (!buf) { + ret = -1; + goto out; + } + ret = dict_set_dynstr (dict, key, buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save snapname (%s) " + "in the dictionary", + snap->snapname); + GF_FREE (buf); + goto out; + } + + buf = NULL; + i++; + } + + ret = dict_set_int32 (dict, "snapcount", i - 1); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not " + "save snapcount in the dictionary"); + goto out; + } + break; + } + + case GF_SNAP_STATUS_TYPE_SNAP: + { + /* IF we give "gluster snapshot status <snapname>" + * then lock is held on single snap. + * This is the place where necessary information + * (snapname)is populated in dictionary + * for locking. + */ + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to fetch snap name"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + snprintf (err_str, len, "Snap (%s)" + "does not exist", snapname); + gf_log(this->name, GF_LOG_ERROR, + "%s", err_str); + ret = -1; + goto out; + } + break; + } + case GF_SNAP_STATUS_TYPE_VOL: + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to fetch volname"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, len, "Volume (%s) " + "does not exist", volname); + gf_log (this->name, GF_LOG_ERROR, + "%s", err_str); + goto out; + } + + i = 1; + list_for_each_entry (snap_volinfo, + &volinfo->snap_volumes, snapvol_list) { + snprintf (key, sizeof (key), "snapname%d", i); + + buf = gf_strdup + (snap_volinfo->snapshot->snapname); + if (!buf) { + ret = -1; + goto out; + } + + ret = dict_set_dynstr (dict, key, buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save snapname"); + GF_FREE (buf); + goto out; + } + + buf = NULL; + i++; + } + + ret = dict_set_int32 (dict, "snapcount", i-1); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save snapcount"); + goto out; + } + break; + default: + { + gf_log (this->name, GF_LOG_ERROR, "Unknown type"); + ret = -1; + goto out; + } + } + + /* Volume lock is not necessary for snapshot status, hence + * turning it off + */ + ret = dict_set_int8 (dict, "hold_vol_locks", 0); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Setting volume lock " + "flag failed"); + goto out; + } + + ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to initiate " + "snap phases"); + goto out; + } + + ret = 0; + +out: + if (voldict) { + dict_unref (voldict); + } + return ret; +} + + +/* This is a snapshot restore handler function. This function will be + * executed in the originator node. This function is responsible for + * calling mgmt_v3 framework to do the actual restore on all the bricks + * + * @param req RPC request object + * @param op gluster operation + * @param dict dictionary containing snapshot restore request + * @param err_str In case of an err this string should be populated + * @param len length of err_str buffer + * + * @return Negative value on Failure and 0 in success + */ +int +glusterd_handle_snapshot_restore (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + char *snapname = NULL; + char *buf = NULL; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + glusterd_snap_t *snap = NULL; + glusterd_volinfo_t *snap_volinfo = NULL; + int32_t i = 0; + char key[PATH_MAX] = ""; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + + GF_ASSERT (conf); + GF_ASSERT (req); + GF_ASSERT (dict); + GF_ASSERT (err_str); + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "get snapname"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + snprintf (err_str, len, "Snap (%s) does not exist", snapname); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + ret = -1; + goto out; + } + + list_for_each_entry (snap_volinfo, &snap->volumes, vol_list) { + i++; + snprintf (key, sizeof (key), "volname%d", i); + buf = gf_strdup (snap_volinfo->parent_volname); + if (!buf) { + ret = -1; + goto out; + } + ret = dict_set_dynstr (dict, key, buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not set " + "parent volume name %s in the dict", + snap_volinfo->parent_volname); + GF_FREE (buf); + goto out; + } + buf = NULL; + } + + ret = dict_set_int32 (dict, "volcount", i); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save volume count"); + goto out; + } + + ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to initiate snap " + "phases"); + goto out; + } + + ret = 0; + +out: + return ret; +} + +glusterd_snap_t* +glusterd_create_snap_object (dict_t *dict, dict_t *rsp_dict) +{ + char *snapname = NULL; + uuid_t *snap_id = NULL; + char *description = NULL; + glusterd_snap_t *snap = NULL; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + int ret = -1; + int64_t time_stamp = 0; + + this = THIS; + priv = this->private; + + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + /* Fetch snapname, description, id and time from dict */ + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snapname"); + goto out; + } + + /* Ignore ret value for description*/ + ret = dict_get_str (dict, "description", &description); + + ret = dict_get_bin (dict, "snap-id", (void **)&snap_id); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snap_id"); + goto out; + } + + ret = dict_get_int64 (dict, "snap-time", &time_stamp); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snap-time"); + goto out; + } + if (time_stamp <= 0) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Invalid time-stamp: %ld", + time_stamp); + goto out; + } + + list_for_each_entry (snap, &priv->snapshots, snap_list) { + if (!strcmp (snap->snapname, snapname) || + !uuid_compare (snap->snap_id, *snap_id)) { + gf_log (THIS->name, GF_LOG_ERROR, + "Found duplicate snap %s (%s)", + snap->snapname, uuid_utoa (snap->snap_id)); + ret = -1; + break; + } + } + if (ret) { + snap = NULL; + goto out; + } + + snap = glusterd_new_snap_object (); + if (!snap) { + gf_log (this->name, GF_LOG_ERROR, "Could not create " + "the snap object for snap %s", snapname); + goto out; + } + + strcpy (snap->snapname, snapname); + uuid_copy (snap->snap_id, *snap_id); + snap->time_stamp = (time_t)time_stamp; + /* Set the status as GD_SNAP_STATUS_INIT and once the backend snapshot + is taken and snap is really ready to use, set the status to + GD_SNAP_STATUS_IN_USE. This helps in identifying the incomplete + snapshots and cleaning them up. + */ + snap->snap_status = GD_SNAP_STATUS_INIT; + if (description) { + snap->description = gf_strdup (description); + if (snap->description == NULL) { + gf_log (this->name, GF_LOG_ERROR, + "Saving the Snap Description Failed"); + ret = -1; + goto out; + } + } + + ret = glusterd_store_snap (snap); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Could not store snap" + "object %s", snap->snapname); + goto out; + } + + list_add_order (&snap->snap_list, &priv->snapshots, + glusterd_compare_snap_time); + + gf_log (this->name, GF_LOG_TRACE, "Snap %s added to the list", + snap->snapname); + + ret = 0; + +out: + if (ret) { + if (snap) + glusterd_snap_remove (rsp_dict, snap, + _gf_true, _gf_true); + snap = NULL; + } + + return snap; +} + +/* This function is called to get the device path of the snap lvm. Usually + if /dev/mapper/<group-name>-<lvm-name> is the device for the lvm, + then the snap device will be /dev/<group-name>/<snapname>. + This function takes care of building the path for the snap device. +*/ +char * +glusterd_build_snap_device_path (char *device, char *snapname) +{ + char snap[PATH_MAX] = ""; + char msg[1024] = ""; + char volgroup[PATH_MAX] = ""; + char *snap_device = NULL; + xlator_t *this = NULL; + runner_t runner = {0,}; + char *ptr = NULL; + int ret = -1; + + this = THIS; + GF_ASSERT (this); + if (!device) { + gf_log (this->name, GF_LOG_ERROR, "device is NULL"); + goto out; + } + if (!snapname) { + gf_log (this->name, GF_LOG_ERROR, "snapname is NULL"); + goto out; + } + + runinit (&runner); + runner_add_args (&runner, "/sbin/lvs", "--noheadings", "-o", "vg_name", + device, NULL); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + snprintf (msg, sizeof (msg), "Get volume group for device %s", device); + runner_log (&runner, this->name, GF_LOG_DEBUG, msg); + ret = runner_start (&runner); + if (ret == -1) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get volume group " + "for device %s", device); + runner_end (&runner); + goto out; + } + ptr = fgets(volgroup, sizeof(volgroup), + runner_chio (&runner, STDOUT_FILENO)); + if (!ptr || !strlen(volgroup)) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get volume group " + "for snap %s", snapname); + runner_end (&runner); + ret = -1; + goto out; + } + runner_end (&runner); + + snprintf (snap, sizeof(snap), "/dev/%s/%s", gf_trim(volgroup), + snapname); + snap_device = gf_strdup (snap); + if (!snap_device) { + gf_log (this->name, GF_LOG_WARNING, "Cannot copy the " + "snapshot device name for snapname: %s)", snapname); + } + +out: + return snap_device; +} + +/* This function actually calls the command (or the API) for taking the + snapshot of the backend brick filesystem. If this is successful, + then call the glusterd_snap_create function to create the snap object + for glusterd +*/ +char * +glusterd_take_lvm_snapshot (glusterd_volinfo_t *snap_vol, + glusterd_brickinfo_t *brickinfo) +{ + char msg[NAME_MAX] = ""; + char buf[PATH_MAX] = ""; + char *snap_device = NULL; + char *ptr = NULL; + char *device = NULL; + int ret = -1; + gf_boolean_t match = _gf_false; + runner_t runner = {0,}; + xlator_t *this = NULL; + + this = THIS; + + if (!brickinfo) { + gf_log (this->name, GF_LOG_ERROR, "brickinfo NULL"); + goto out; + } + + device = glusterd_get_brick_mount_details (brickinfo); + if (!device) { + gf_log (this->name, GF_LOG_ERROR, "getting device name for " + "the brick %s:%s failed", brickinfo->hostname, + brickinfo->path); + goto out; + } + + /* Figuring out if setactivationskip flag is supported or not */ + runinit (&runner); + snprintf (msg, sizeof (msg), "running lvcreate help"); + runner_add_args (&runner, "/sbin/lvcreate", "--help", NULL); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + ret = runner_start (&runner); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to run lvcreate help"); + runner_end (&runner); + goto out; + } + + /* Looking for setactivationskip in lvcreate --help */ + do { + ptr = fgets(buf, sizeof(buf), + runner_chio (&runner, STDOUT_FILENO)); + if (ptr) { + if (strstr(buf, "setactivationskip")) { + match = _gf_true; + break; + } + } + } while (ptr != NULL); + runner_end (&runner); + + /* Takng the actual snapshot */ + runinit (&runner); + snprintf (msg, sizeof (msg), "taking snapshot of the brick %s:%s", + brickinfo->hostname, brickinfo->path); + if (match == _gf_true) + runner_add_args (&runner, "/sbin/lvcreate", "-s", device, + "--setactivationskip", "n", "--name", + snap_vol->volname, NULL); + else + runner_add_args (&runner, "/sbin/lvcreate", "-s", device, + "--name", snap_vol->volname, NULL); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + ret = runner_start (&runner); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "taking snapshot of the " + "brick (%s:%s) of device %s failed", + brickinfo->hostname, brickinfo->path, device); + runner_end (&runner); + goto out; + } + runner_end (&runner); + + snap_device = glusterd_build_snap_device_path (device, + snap_vol->volname); + if (!snap_device) { + gf_log (this->name, GF_LOG_WARNING, "Cannot copy the snapshot " + "device name for snap %s (volume id: %s)", + snap_vol->snapshot->snapname, snap_vol->volname); + ret = -1; + goto out; + } + +out: + return snap_device; +} + +int32_t +glusterd_snap_brick_create (char *device, glusterd_volinfo_t *snap_volinfo, + glusterd_brickinfo_t *original_brickinfo, + int32_t brick_count, char *snap_brick_dir) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + char snap_brick_mount_path[PATH_MAX] = ""; + char snap_brick_path[PATH_MAX] = ""; + char msg[1024] = ""; + struct stat statbuf = {0, }; + runner_t runner = {0, }; + + this = THIS; + priv = this->private; + + GF_ASSERT (device); + GF_ASSERT (snap_volinfo); + GF_ASSERT (original_brickinfo); + GF_ASSERT (snap_brick_dir); + + snprintf (snap_brick_mount_path, sizeof (snap_brick_mount_path), + "%s/%s/brick%d", snap_mount_folder, snap_volinfo->volname, + brick_count+1); + + snprintf (snap_brick_path, sizeof (snap_brick_path), "%s%s", + snap_brick_mount_path, snap_brick_dir); + + ret = mkdir_p (snap_brick_mount_path, 0777, _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "creating the brick directory" + " %s for the snapshot %s(device: %s) failed", + snap_brick_mount_path, snap_volinfo->volname, device); + goto out; + } + /* mount the snap logical device on the directory inside + /run/gluster/snaps/<snapname>/@snap_brick_mount_path + Way to mount the snap brick via mount api is this. + ret = mount (device, snap_brick_mount_path, entry->mnt_type, + MS_MGC_VAL, "nouuid"); + But for now, mounting using runner apis. + */ + runinit (&runner); + snprintf (msg, sizeof (msg), "mounting snapshot of the brick %s:%s", + original_brickinfo->hostname, original_brickinfo->path); + runner_add_args (&runner, "mount", "-o", "nouuid", device, + snap_brick_mount_path, NULL); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + + /* let glusterd get blocked till snapshot is over */ + synclock_unlock (&priv->big_lock); + ret = runner_run (&runner); + synclock_lock (&priv->big_lock); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "mounting the snapshot " + "logical device %s failed (error: %s)", device, + strerror (errno)); + goto out; + } else + gf_log (this->name, GF_LOG_DEBUG, "mounting the snapshot " + "logical device %s successful", device); + + ret = stat (snap_brick_path, &statbuf); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "stat of the brick %s" + "(brick mount: %s) failed (%s)", snap_brick_path, + snap_brick_mount_path, strerror (errno)); + goto out; + } + ret = sys_lsetxattr (snap_brick_path, + GF_XATTR_VOL_ID_KEY, + snap_volinfo->volume_id, 16, + XATTR_REPLACE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "extended attribute %s on %s. Reason: " + "%s, snap: %s", GF_XATTR_VOL_ID_KEY, + snap_brick_path, strerror (errno), + snap_volinfo->volname); + goto out; + } + +out: + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "unmounting the snap brick" + " mount %s", snap_brick_mount_path); + umount (snap_brick_mount_path); + } + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Added missed_snap_entry to rsp_dict */ +int32_t +glusterd_add_missed_snaps_to_dict (dict_t *rsp_dict, char *snap_uuid, + glusterd_brickinfo_t *brickinfo, + int32_t brick_number, int32_t op) +{ + char *buf = NULL; + char missed_snap_entry[PATH_MAX] = ""; + char name_buf[PATH_MAX] = ""; + int32_t missed_snap_count = -1; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap_uuid); + GF_ASSERT (brickinfo); + + snprintf (missed_snap_entry, sizeof(missed_snap_entry), + "%s:%s=%d:%s:%d:%d", uuid_utoa(brickinfo->uuid), + snap_uuid, brick_number, brickinfo->path, op, + GD_MISSED_SNAP_PENDING); + + buf = gf_strdup (missed_snap_entry); + if (!buf) { + ret = -1; + goto out; + } + + /* Fetch the missed_snap_count from the dict */ + ret = dict_get_int32 (rsp_dict, "missed_snap_count", + &missed_snap_count); + if (ret) { + /* Initialize the missed_snap_count for the first time */ + missed_snap_count = 0; + } + + /* Setting the missed_snap_entry in the rsp_dict */ + snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d", + missed_snap_count); + ret = dict_set_dynstr (rsp_dict, name_buf, buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set missed_snap_entry (%s) " + "in the rsp_dict.", buf); + GF_FREE (buf); + goto out; + } + missed_snap_count++; + + /* Setting the new missed_snap_count in the dict */ + ret = dict_set_int32 (rsp_dict, "missed_snap_count", + missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set missed_snap_count for %s " + "in the rsp_dict.", missed_snap_entry); + goto out; + } + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int32_t +glusterd_add_bricks_to_snap_volume (dict_t *dict, dict_t *rsp_dict, + glusterd_volinfo_t *snap_vol, + glusterd_brickinfo_t *original_brickinfo, + glusterd_brickinfo_t *snap_brickinfo, + char **snap_brick_dir, int64_t volcount, + int32_t brick_count) +{ + char key[PATH_MAX] = ""; + char snap_brick_path[PATH_MAX] = ""; + char *snap_device = NULL; + gf_boolean_t add_missed_snap = _gf_false; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap_vol); + GF_ASSERT (original_brickinfo); + GF_ASSERT (snap_brickinfo); + GF_ASSERT (snap_brick_dir); + + snprintf (key, sizeof(key) - 1, "vol%ld.brickdir%d", volcount, + brick_count); + ret = dict_get_ptr (dict, key, (void **)snap_brick_dir); + if (ret) { + /* Using original brickinfo here because it will be a + * pending snapshot and storing the original brickinfo + * will help in mapping while recreating the missed snapshot + */ + gf_log (this->name, GF_LOG_WARNING, "Unable to fetch " + "snap mount path (%s). Using original brickinfo", key); + snap_brickinfo->snap_status = -1; + strcpy (snap_brick_path, original_brickinfo->path); + + /* In origiator node add snaps missed + * from different nodes to the dict + */ + if (is_origin_glusterd (dict) == _gf_true) + add_missed_snap = _gf_true; + } else { + /* Create brick-path in the format /var/run/gluster/snaps/ * + * <snap-uuid>/<original-brick#>/snap-brick-dir * + */ + snprintf (snap_brick_path, sizeof(snap_brick_path), + "%s/%s/brick%d%s", snap_mount_folder, + snap_vol->volname, brick_count+1, + *snap_brick_dir); + } + + if ((snap_brickinfo->snap_status != -1) && + (!uuid_compare (original_brickinfo->uuid, MY_UUID)) && + (!glusterd_is_brick_started (original_brickinfo))) { + /* In case if the brick goes down after prevalidate. */ + gf_log (this->name, GF_LOG_WARNING, "brick %s:%s is not" + " started (snap: %s)", + original_brickinfo->hostname, + original_brickinfo->path, + snap_vol->snapshot->snapname); + + snap_brickinfo->snap_status = -1; + strcpy (snap_brick_path, original_brickinfo->path); + add_missed_snap = _gf_true; + } + + if (add_missed_snap) { + ret = glusterd_add_missed_snaps_to_dict (rsp_dict, + snap_vol->volname, + original_brickinfo, + brick_count + 1, + GF_SNAP_OPTION_TYPE_CREATE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to add missed" + " snapshot info for %s:%s in the rsp_dict", + original_brickinfo->hostname, + original_brickinfo->path); + goto out; + } + } + + snprintf (key, sizeof(key), "vol%ld.brick_snapdevice%d", + volcount, brick_count); + ret = dict_get_ptr (dict, key, (void **)&snap_device); + if (ret) { + /* If the device name is empty, so will be the brick path + * Hence the missed snap has already been added above + */ + gf_log (this->name, GF_LOG_ERROR, "Unable to fetch " + "snap device (%s). Leaving empty", key); + } else + strcpy (snap_brickinfo->device_path, snap_device); + + ret = gf_canonicalize_path (snap_brick_path); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to canonicalize path"); + goto out; + } + + strcpy (snap_brickinfo->hostname, original_brickinfo->hostname); + strcpy (snap_brickinfo->path, snap_brick_path); + uuid_copy (snap_brickinfo->uuid, original_brickinfo->uuid); + list_add_tail (&snap_brickinfo->brick_list, &snap_vol->bricks); + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +static int32_t +glusterd_take_brick_snapshot (glusterd_volinfo_t *origin_vol, + glusterd_volinfo_t *snap_vol, dict_t *rsp_dict, + glusterd_brickinfo_t *original_brickinfo, + glusterd_brickinfo_t *snap_brickinfo, + char *snap_brick_dir, int32_t brick_count) +{ + char *device = NULL; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (origin_vol); + GF_ASSERT (snap_vol); + GF_ASSERT (rsp_dict); + GF_ASSERT (original_brickinfo); + GF_ASSERT (snap_brickinfo); + GF_ASSERT (snap_brick_dir); + + device = glusterd_take_lvm_snapshot (snap_vol, original_brickinfo); + /* Fail the snapshot even though snapshot on one of + the bricks fails. At the end when we check whether + the snapshot volume meets quorum or not, then the + the snapshot can either be treated as success, or + in case of failure we can undo the changes and return + failure to cli. */ + if (!device) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to take snapshot of %s:%s", + original_brickinfo->hostname, + original_brickinfo->path); + goto out; + } + + /* create the complete brick here */ + ret = glusterd_snap_brick_create (device, snap_vol, + original_brickinfo, + brick_count, snap_brick_dir); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "not able to" + " create the brickinfo for the snap %s" + ", volume %s", snap_vol->snapshot->snapname, + origin_vol->volname); + goto out; + } + +out: + if (device) + GF_FREE (device); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Look for disconnected peers, for missed snap creates or deletes */ +static int32_t +glusterd_find_missed_snap (dict_t *rsp_dict, glusterd_volinfo_t *vol, + char *snap_uuid, struct list_head *peers, + int32_t op) +{ + int32_t brick_count = -1; + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (peers); + GF_ASSERT (vol); + GF_ASSERT (snap_uuid); + + brick_count = 0; + list_for_each_entry (brickinfo, &vol->bricks, brick_list) { + if (!uuid_compare (brickinfo->uuid, MY_UUID)) { + /* If the brick belongs to the same node */ + brick_count++; + continue; + } + + list_for_each_entry (peerinfo, peers, uuid_list) { + if (uuid_compare (peerinfo->uuid, brickinfo->uuid)) { + /* If the brick doesnt belong to this peer */ + continue; + } + + /* Found peer who owns the brick, * + * if peer is not connected or not * + * friend add it to missed snap list */ + if (!(peerinfo->connected) || + (peerinfo->state.state != + GD_FRIEND_STATE_BEFRIENDED)) { + ret = glusterd_add_missed_snaps_to_dict + (rsp_dict, + snap_uuid, + brickinfo, + brick_count + 1, + op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add missed snapshot " + "info for %s:%s in the " + "rsp_dict", brickinfo->hostname, + brickinfo->path); + goto out; + } + } + } + brick_count++; + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +glusterd_volinfo_t * +glusterd_do_snap_vol (glusterd_volinfo_t *origin_vol, glusterd_snap_t *snap, + dict_t *dict, dict_t *rsp_dict, int64_t volcount) +{ + char key[PATH_MAX] = ""; + char *snap_brick_dir = NULL; + char *username = NULL; + char *password = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_conf_t *priv = NULL; + glusterd_volinfo_t *snap_vol = NULL; + uuid_t *snap_volid = NULL; + int32_t ret = -1; + int32_t brick_count = 0; + glusterd_brickinfo_t *snap_brickinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (origin_vol); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + /* fetch username, password and vol_id from dict*/ + snprintf (key, sizeof(key), "volume%ld_username", volcount); + ret = dict_get_str (dict, key, &username); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get %s for " + "snap %s", key, snap->snapname); + goto out; + } + + snprintf (key, sizeof(key), "volume%ld_password", volcount); + ret = dict_get_str (dict, key, &password); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get %s for " + "snap %s", key, snap->snapname); + goto out; + } + snprintf (key, sizeof(key) - 1, "vol%ld_volid", volcount); + ret = dict_get_bin (dict, key, (void **)&snap_volid); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch snap_volid"); + goto out; + } + + /* We are not setting the username and password here as + * we need to set the user name and password passed in + * the dictionary + */ + ret = glusterd_volinfo_dup (origin_vol, &snap_vol, _gf_false); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to duplicate volinfo " + "for the snapshot %s", snap->snapname); + goto out; + } + + /* uuid is used as lvm snapshot name. + This will avoid restrictions on snapshot names provided by user */ + GLUSTERD_GET_UUID_NOHYPHEN (snap_vol->volname, *snap_volid); + uuid_copy (snap_vol->volume_id, *snap_volid); + snap_vol->is_snap_volume = _gf_true; + strcpy (snap_vol->parent_volname, origin_vol->volname); + snap_vol->snapshot = snap; + + glusterd_auth_set_username (snap_vol, username); + glusterd_auth_set_password (snap_vol, password); + + /* Adding snap brickinfos to the snap volinfo */ + brick_count = 0; + list_for_each_entry (brickinfo, &origin_vol->bricks, brick_list) { + snap_brickinfo = NULL; + + ret = glusterd_brickinfo_new (&snap_brickinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "initializing the brick for the snap " + "volume failed (snapname: %s)", snap->snapname); + goto out; + } + + ret = glusterd_add_bricks_to_snap_volume (dict, rsp_dict, + snap_vol, + brickinfo, + snap_brickinfo, + &snap_brick_dir, + volcount, + brick_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add the snap brick for " + "%s:%s to the snap volume", + brickinfo->hostname, brickinfo->path); + GF_FREE (snap_brickinfo); + goto out; + } + + /* Take snapshot of the brick */ + if ((uuid_compare (brickinfo->uuid, MY_UUID)) || + (snap_brickinfo->snap_status == -1)) { + brick_count++; + continue; + } + + ret = glusterd_take_brick_snapshot (origin_vol, snap_vol, + rsp_dict, brickinfo, + snap_brickinfo, + snap_brick_dir, + brick_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to take snapshot for %s:%s", + brickinfo->hostname, brickinfo->path); + goto out; + } + + brick_count++; + } + + /*TODO: the quorum check of the snap volume here */ + + ret = glusterd_store_volinfo (snap_vol, + GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to store snapshot " + "volinfo (%s) for snap %s", snap_vol->volname, + snap->snapname); + goto out; + } + + ret = generate_brick_volfiles (snap_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "generating the brick " + "volfiles for the snap %s (volume: %s) failed", + snap->snapname, origin_vol->volname); + goto out; + } + + ret = generate_client_volfiles (snap_vol, GF_CLIENT_TRUSTED); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "generating the trusted " + "client volfiles for the snap %s (volume: %s) failed", + snap->snapname, origin_vol->volname); + goto out; + } + ret = generate_client_volfiles (snap_vol, GF_CLIENT_OTHER); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "generating the client " + "volfiles for the snap %s (volume: %s) failed", + snap->snapname, origin_vol->volname); + goto out; + } + + ret = glusterd_list_add_snapvol (origin_vol, snap_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "could not add the snap " + "volume %s to the list", snap_vol->volname); + goto out; + } + + list_for_each_entry (brickinfo, &snap_vol->bricks, brick_list) { + if (uuid_compare (brickinfo->uuid, MY_UUID)) + continue; + + if (brickinfo->snap_status == -1) { + gf_log (this->name, GF_LOG_INFO, + "not starting snap brick %s:%s for " + "for the snap %s (volume: %s)", + brickinfo->hostname, brickinfo->path, + snap->snapname, origin_vol->volname); + continue; + } + + ret = glusterd_brick_start (snap_vol, brickinfo, _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "starting the " + "brick %s:%s for the snap %s (volume: %s) " + "failed", brickinfo->hostname, brickinfo->path, + snap->snapname, origin_vol->volname); + goto out; + } + } + + snap_vol->status = GLUSTERD_STATUS_STARTED; + ret = glusterd_store_volinfo (snap_vol, + GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to store snap volinfo"); + goto out; + } + +out: + if (ret) { + if (snap_vol) + glusterd_snap_volume_remove (rsp_dict, snap_vol, + _gf_true, _gf_true); + snap_vol = NULL; + } + + return snap_vol; +} + +/* This is a snapshot remove handler function. This function will be + * executed in the originator node. This function is responsible for + * calling mgmt v3 framework to do the actual remove on all the bricks + * + * @param req RPC request object + * @param op gluster operation + * @param dict dictionary containing snapshot remove request + * @param err_str In case of an err this string should be populated + * @param len length of err_str buffer + * + * @return Negative value on Failure and 0 in success + */ +int +glusterd_handle_snapshot_remove (rpcsvc_request_t *req, glusterd_op_t op, + dict_t *dict, char *err_str, size_t len) +{ + int ret = -1; + int64_t volcount = 0; + char *snapname = NULL; + char *volname = NULL; + char key[PATH_MAX] = ""; + glusterd_snap_t *snap = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_volinfo_t *tmp = NULL; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (req); + GF_ASSERT (dict); + GF_ASSERT (err_str); + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get snapname"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + snprintf (err_str, len, "Snap (%s) does not exist", snapname); + gf_log (this->name, GF_LOG_ERROR, + "%s", err_str); + ret = -1; + goto out; + } + + /* Set volnames in the dict to get mgmt_v3 lock */ + list_for_each_entry_safe (snap_vol, tmp, &snap->volumes, vol_list) { + volcount++; + volname = gf_strdup (snap_vol->parent_volname); + if (!volname) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "strdup failed"); + goto out; + } + + snprintf (key, sizeof (key), "volname%ld", volcount); + ret = dict_set_dynstr (dict, key, volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set " + "volume name in dictionary"); + GF_FREE (volname); + goto out; + } + volname = NULL; + } + ret = dict_set_int64 (dict, "volcount", volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set volcount"); + goto out; + } + + ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to initiate snap " + "phases"); + goto out; + } + + ret = 0; +out: + return ret; +} + +int +glusterd_snapshot_remove_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int32_t ret = -1; + char *snapname = NULL; + xlator_t *this = NULL; + glusterd_snap_t *snap = NULL; + + this = THIS; + + if (!dict || !op_errstr) { + gf_log (this->name, GF_LOG_ERROR, "input parameters NULL"); + goto out; + } + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Getting the snap name " + "failed"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + gf_log (this->name, GF_LOG_ERROR, "Snap %s does not exist", + snapname); + ret = -1; + goto out; + } + ret = 0; +out: + return ret; +} + +int +glusterd_snapshot_status_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int ret = -1; + char *snapname = NULL; + glusterd_conf_t *conf = NULL; + xlator_t *this = NULL; + int32_t cmd = -1; + glusterd_volinfo_t *volinfo = NULL; + char *volname = NULL; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + + GF_ASSERT (conf); + GF_ASSERT (op_errstr); + if (!dict) { + gf_log (this->name, GF_LOG_ERROR, "Input dict is NULL"); + goto out; + } + + ret = dict_get_int32 (dict, "cmd", &cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not fetch status cmd"); + goto out; + } + + switch (cmd) { + case GF_SNAP_STATUS_TYPE_ALL: + { + break; + } + case GF_SNAP_STATUS_TYPE_SNAP: + { + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not fetch snapname"); + goto out; + } + + if (!glusterd_find_snap_by_name (snapname)) { + ret = gf_asprintf (op_errstr, "Snap (%s) " + "not found", snapname); + if (ret < 0) { + goto out; + } + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Snap (%s) " + "not found", snapname); + goto out; + } + break; + } + case GF_SNAP_STATUS_TYPE_VOL: + { + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not fetch volname"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + ret = gf_asprintf (op_errstr, "Volume (%s)" + "not found", volname); + if (ret < 0) { + goto out; + } + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Volume " + "%s not present", volname); + goto out; + } + break; + + } + default: + { + gf_log (this->name, GF_LOG_ERROR, "Invalid command"); + break; + } + } + ret = 0; + +out: + return ret; +} + +int32_t +glusterd_snapshot_remove_commit (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int32_t ret = -1; + char *snapname = NULL; + char *dup_snapname = NULL; + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + glusterd_volinfo_t *snap_volinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + GF_ASSERT (op_errstr); + + priv = this->private; + GF_ASSERT (priv); + + if (!dict || !op_errstr) { + gf_log (this->name, GF_LOG_ERROR, "input parameters NULL"); + goto out; + } + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Getting the snap name " + "failed"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + gf_log (this->name, GF_LOG_ERROR, "Snap %s does not exist", + snapname); + ret = -1; + goto out; + } + + if (is_origin_glusterd (dict) == _gf_true) { + /* TODO : As of now there is only volume in snapshot. + * Change this when multiple volume snapshot is introduced + */ + snap_volinfo = list_entry (snap->volumes.next, + glusterd_volinfo_t, + vol_list); + if (!snap_volinfo) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch snap_volinfo"); + ret = -1; + goto out; + } + + /* From origin glusterd check if * + * any peers with snap bricks is down */ + ret = glusterd_find_missed_snap (rsp_dict, snap_volinfo, + snap_volinfo->volname, + &priv->peers, + GF_SNAP_OPTION_TYPE_DELETE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to find missed snap deletes"); + goto out; + } + } + + ret = glusterd_snap_remove (rsp_dict, snap, _gf_true, _gf_false); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to remove snap %s", + snapname); + goto out; + } + + dup_snapname = gf_strdup (snapname); + if (!dup_snapname) { + gf_log (this->name, GF_LOG_ERROR, "Strdup failed"); + ret = -1; + goto out; + } + + ret = dict_set_dynstr (rsp_dict, "snapname", dup_snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to set the snapname"); + GF_FREE (dup_snapname); + goto out; + } + + ret = 0; +out: + return ret; +} + +int32_t +glusterd_do_snap_cleanup (dict_t *dict, char **op_errstr, dict_t *rsp_dict) +{ + int32_t ret = -1; + char *name = NULL; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + glusterd_snap_t *snap = NULL; + + this = THIS; + GF_ASSERT (this); + conf = this->private; + GF_ASSERT (conf); + + if (!dict || !op_errstr) { + gf_log (this->name, GF_LOG_ERROR, "input parameters NULL"); + goto out; + } + + ret = dict_get_str (dict, "snapname", &name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "getting the snap " + "name failed (volume: %s)", volinfo->volname); + goto out; + } + + /* + If the snapname is not found that means the failure happened at + staging, or in commit, before the snap object is created, in which + case there is nothing to cleanup. So set ret to 0. + */ + snap = glusterd_find_snap_by_name (name); + if (!snap) { + gf_log (this->name, GF_LOG_INFO, "snap %s is not found", name); + ret = 0; + goto out; + } + + ret = glusterd_snap_remove (rsp_dict, snap, _gf_true, _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "removing the snap %s failed", + name); + goto out; + } + + name = NULL; + + ret = 0; + +out: + + return ret; +} + +/* In case of a successful, delete or create operation, during post_validate * + * look for missed snap operations and update the missed snap lists */ +int32_t +glusterd_snapshot_update_snaps_post_validate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int32_t ret = -1; + int32_t missed_snap_count = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + GF_ASSERT (op_errstr); + + ret = dict_get_int32 (dict, "missed_snap_count", + &missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "No missed snaps"); + ret = 0; + goto out; + } + + ret = glusterd_store_update_missed_snaps (dict, missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to update missed_snaps_list"); + goto out; + } + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_snapshot_create_commit (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int ret = -1; + int64_t i = 0; + int64_t volcount = 0; + char *snapname = NULL; + char *volname = NULL; + char *tmp_name = NULL; + char key[PATH_MAX] = ""; + xlator_t *this = NULL; + glusterd_snap_t *snap = NULL; + glusterd_volinfo_t *origin_vol = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_conf_t *priv = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(dict); + GF_ASSERT(op_errstr); + GF_ASSERT(rsp_dict); + priv = this->private; + GF_ASSERT(priv); + + ret = dict_get_int64 (dict, "volcount", &volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "get the volume count"); + goto out; + } + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snapname"); + goto out; + } + tmp_name = gf_strdup (snapname); + if (!tmp_name) { + gf_log (this->name, GF_LOG_ERROR, "Out of memory"); + ret = -1; + goto out; + } + + ret = dict_set_dynstr (rsp_dict, "snapname", tmp_name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set snapname in rsp_dict"); + GF_FREE (tmp_name); + goto out; + } + tmp_name = NULL; + + snap = glusterd_create_snap_object (dict, rsp_dict); + if (!snap) { + gf_log (this->name, GF_LOG_ERROR, "creating the" + "snap object %s failed", snapname); + ret = -1; + goto out; + } + + for (i = 1; i <= volcount; i++) { + snprintf (key, sizeof (key), "volname%ld", i); + ret = dict_get_str (dict, key, &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get volume name"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &origin_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get the volinfo for " + "the volume %s", volname); + goto out; + } + + /* TODO: Create a stub where the bricks are + added parallely by worker threads so that + the snap creating happens parallely. */ + snap_vol = glusterd_do_snap_vol (origin_vol, snap, dict, + rsp_dict, i); + if (!snap_vol) { + ret = -1; + gf_log (this->name, GF_LOG_WARNING, "taking the " + "snapshot of the volume %s failed", volname); + goto out; + } + } + + snap->snap_status = GD_SNAP_STATUS_IN_USE; + ret = glusterd_store_snap (snap); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Could not store snap" + "object %s", snap->snapname); + goto out; + } + + ret = 0; + +out: + if (ret) { + if (snap) + glusterd_snap_remove (rsp_dict, snap, + _gf_true, _gf_true); + snap = NULL; + } + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +snap_max_hard_limit_set_commit (dict_t *dict, uint64_t value, + char *volname, char **op_errstr) +{ + char err_str[PATH_MAX] = ""; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = -1; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (volname); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + + /* TODO: Initiate auto deletion when there is a limit change */ + if (!volname) { + /* For system limit */ + conf->snap_max_hard_limit = value; + + ret = glusterd_store_global_info (this); + if (ret) { + snprintf (err_str, PATH_MAX, "Failed to store " + "snap-max-hard-limit for system"); + goto out; + } + } else { + /* For one volume */ + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, PATH_MAX, "Failed to get the" + " volinfo for volume %s", volname); + goto out; + } + + volinfo->snap_max_hard_limit = value; + + ret = glusterd_store_volinfo (volinfo, + GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + snprintf (err_str, PATH_MAX, "Failed to store " + "snap-max-hard-limit for volume %s", volname); + goto out; + } + } + + ret = 0; +out: + if (ret) { + *op_errstr = gf_strdup (err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + } + return ret; +} + +int +snap_max_limits_display_commit (dict_t *rsp_dict, char *volname, + char **op_errstr) +{ + char err_str[PATH_MAX] = ""; + char buf[PATH_MAX] = ""; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *volinfo = NULL; + int ret = -1; + uint64_t active_hard_limit = 0; + uint64_t snap_max_limit = 0; + uint64_t soft_limit_value = -1; + uint64_t count = 0; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + GF_ASSERT (volname); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + + if (!volname) { + /* For system limit */ + list_for_each_entry (volinfo, &conf->volumes, vol_list) { + if (volinfo->is_snap_volume == _gf_true) + continue; + snap_max_limit = volinfo->snap_max_hard_limit; + if (snap_max_limit > conf->snap_max_hard_limit) + active_hard_limit = conf->snap_max_hard_limit; + else + active_hard_limit = snap_max_limit; + soft_limit_value = (active_hard_limit * + conf->snap_max_soft_limit) / 100; + + snprintf (buf, sizeof(buf), "volume%ld-volname", count); + ret = dict_set_str (rsp_dict, buf, volinfo->volname); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-hard-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, snap_max_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-active-hard-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, + active_hard_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-soft-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, soft_limit_value); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + count++; + } + + ret = dict_set_uint64 (rsp_dict, "voldisplaycount", count); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set voldisplaycount"); + goto out; + } + } else { + /* For one volume */ + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + snprintf (err_str, PATH_MAX, "Failed to get the" + " volinfo for volume %s", volname); + goto out; + } + + snap_max_limit = volinfo->snap_max_hard_limit; + if (snap_max_limit > conf->snap_max_hard_limit) + active_hard_limit = conf->snap_max_hard_limit; + else + active_hard_limit = snap_max_limit; + + soft_limit_value = (active_hard_limit * + conf->snap_max_soft_limit) / 100; + + snprintf (buf, sizeof(buf), "volume%ld-volname", count); + ret = dict_set_str (rsp_dict, buf, volinfo->volname); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-hard-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, snap_max_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-active-hard-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, active_hard_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-soft-limit", count); + ret = dict_set_uint64 (rsp_dict, buf, soft_limit_value); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set %s", buf); + goto out; + } + + count++; + + ret = dict_set_uint64 (rsp_dict, "voldisplaycount", count); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set voldisplaycount"); + goto out; + } + + } + + ret = dict_set_uint64 (rsp_dict, "snap-max-hard-limit", + conf->snap_max_hard_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set sys-snap-max-hard-limit "); + goto out; + } + + ret = dict_set_uint64 (rsp_dict, "snap-max-soft-limit", + conf->snap_max_soft_limit); + if (ret) { + snprintf (err_str, PATH_MAX, + "Failed to set sys-snap-max-hard-limit "); + goto out; + } + + ret = 0; +out: + if (ret) { + *op_errstr = gf_strdup (err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + } + return ret; +} + +int +glusterd_snapshot_config_commit (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + char *volname = NULL; + xlator_t *this = NULL; + int ret = -1; + char err_str[PATH_MAX] = {0,}; + glusterd_conf_t *conf = NULL; + int config_command = 0; + uint64_t hard_limit = 0; + uint64_t soft_limit = 0; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + + ret = dict_get_int32 (dict, "config-command", &config_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to get config-command type"); + goto out; + } + + /* Ignore the return value of the following dict_get, + * as they are optional + */ + ret = dict_get_str (dict, "volname", &volname); + + ret = dict_get_uint64 (dict, "snap-max-hard-limit", &hard_limit); + + ret = dict_get_uint64 (dict, "snap-max-soft-limit", &soft_limit); + + switch (config_command) { + case GF_SNAP_CONFIG_TYPE_SET: + if (hard_limit) { + /* Commit ops for snap-max-hard-limit */ + ret = snap_max_hard_limit_set_commit (dict, hard_limit, + volname, + op_errstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "snap-max-hard-limit set " + "commit failed."); + goto out; + } + } + + if (soft_limit) { + /* For system limit */ + conf->snap_max_soft_limit = soft_limit; + + ret = glusterd_store_global_info (this); + if (ret) { + snprintf (err_str, PATH_MAX, "Failed to store " + "snap-max-soft-limit for system"); + *op_errstr = gf_strdup (err_str); + gf_log (this->name, GF_LOG_ERROR, "%s", + err_str); + goto out; + } + } + break; + + case GF_SNAP_CONFIG_DISPLAY: + /* Reading data from local node only */ + if (!is_origin_glusterd (dict)) { + ret = 0; + break; + } + + ret = snap_max_limits_display_commit (rsp_dict, volname, + op_errstr); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "snap-max-limit " + "display commit failed."); + goto out; + } + break; + default: + break; + } + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_get_brick_lvm_details (dict_t *rsp_dict, + glusterd_brickinfo_t *brickinfo, char *volname, + char *device, char *key_prefix) +{ + + int ret = -1; + glusterd_conf_t *priv = NULL; + runner_t runner = {0,}; + xlator_t *this = NULL; + char msg[PATH_MAX] = ""; + char buf[PATH_MAX] = ""; + char *ptr = NULL; + char *token = NULL; + char key[PATH_MAX] = ""; + char *value = NULL; + + GF_ASSERT (rsp_dict); + GF_ASSERT (brickinfo); + GF_ASSERT (volname); + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + device = glusterd_get_brick_mount_details (brickinfo); + if (!device) { + gf_log (this->name, GF_LOG_ERROR, "Getting device name for " + "the brick %s:%s failed", brickinfo->hostname, + brickinfo->path); + goto out; + } + runinit (&runner); + snprintf (msg, sizeof (msg), "running lvs command, " + "for getting snap status"); + /* Using lvs command fetch the Volume Group name, + * Percentage of data filled and Logical Volume size + * + * "-o" argument is used to get the desired information, + * example : "lvs /dev/VolGroup/thin_vol -o vgname,lv_size", + * will get us Volume Group name and Logical Volume size. + * + * Here separator used is ":", + * for the above given command with separator ":", + * The output will be "vgname:lvsize" + */ + runner_add_args (&runner, "lvs", device, "--noheading", "-o", + "vg_name,data_percent,lv_size", + "--separator", ":", NULL); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + runner_log (&runner, "", GF_LOG_DEBUG, msg); + ret = runner_start (&runner); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not perform lvs action"); + goto end; + } + do { + ptr = fgets (buf, sizeof (buf), + runner_chio (&runner, STDOUT_FILENO)); + + if (ptr == NULL) + break; + token = strtok (buf, ":"); + if (token != NULL) { + while (token && token[0] == ' ') + token++; + if (!token) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, + "Invalid vg entry"); + goto end; + } + value = gf_strdup (token); + if (!value) { + ret = -1; + goto end; + } + ret = snprintf (key, sizeof (key), "%s.vgname", + key_prefix); + if (ret < 0) { + goto end; + } + + ret = dict_set_dynstr (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save vgname "); + goto end; + } + } + + token = strtok (NULL, ":"); + if (token != NULL) { + value = gf_strdup (token); + if (!value) { + ret = -1; + goto end; + } + ret = snprintf (key, sizeof (key), "%s.data", + key_prefix); + if (ret < 0) { + goto end; + } + + ret = dict_set_dynstr (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save data percent "); + goto end; + } + } + token = strtok (NULL, ":"); + if (token != NULL) { + value = gf_strdup (token); + if (!value) { + ret = -1; + goto end; + } + ret = snprintf (key, sizeof (key), "%s.lvsize", + key_prefix); + if (ret < 0) { + goto end; + } + + ret = dict_set_dynstr (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save meta data percent "); + goto end; + } + } + + } while (ptr != NULL); + + ret = 0; + +end: + runner_end (&runner); + +out: + if (ret && value) { + GF_FREE (value); + } + + return ret; +} + +int +glusterd_get_single_brick_status (char **op_errstr, dict_t *rsp_dict, + char *keyprefix, int index, + glusterd_volinfo_t *snap_volinfo, + glusterd_brickinfo_t *brickinfo) +{ + int ret = -1; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + char key[PATH_MAX] = ""; + char *device = NULL; + char *value = NULL; + char brick_path[PATH_MAX] = ""; + char pidfile[PATH_MAX] = ""; + pid_t pid = -1; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + GF_ASSERT (keyprefix); + GF_ASSERT (snap_volinfo); + GF_ASSERT (brickinfo); + + ret = snprintf (key, sizeof (key), "%s.brick%d.path", keyprefix, + index); + if (ret < 0) { + goto out; + } + + ret = snprintf (brick_path, sizeof (brick_path), + "%s:%s", brickinfo->hostname, brickinfo->path); + if (ret < 0) { + goto out; + } + + value = gf_strdup (brick_path); + if (!value) { + ret = -1; + goto out; + } + + ret = dict_set_dynstr (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to store " + "brick_path %s", brickinfo->path); + goto out; + } + + if (brickinfo->snap_status == -1) { + /* Setting vgname as "Pending Snapshot" */ + value = gf_strdup ("Pending Snapshot"); + if (!value) { + ret = -1; + goto out; + } + + snprintf (key, sizeof (key), "%s.brick%d.vgname", + keyprefix, index); + ret = dict_set_dynstr (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save vgname "); + goto out; + } + + ret = 0; + goto out; + } + value = NULL; + + ret = snprintf (key, sizeof (key), "%s.brick%d.status", + keyprefix, index); + if (ret < 0) { + goto out; + } + + if (brickinfo->status == GF_BRICK_STOPPED) { + value = gf_strdup ("No"); + if (!value) { + ret = -1; + goto out; + } + ret = dict_set_str (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save brick status"); + goto out; + } + value = NULL; + } else { + value = gf_strdup ("Yes"); + if (!value) { + ret = -1; + goto out; + } + ret = dict_set_str (rsp_dict, key, value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save brick status"); + goto out; + } + value = NULL; + + GLUSTERD_GET_BRICK_PIDFILE (pidfile, snap_volinfo, + brickinfo, priv); + ret = glusterd_is_service_running (pidfile, &pid); + + ret = snprintf (key, sizeof (key), "%s.brick%d.pid", + keyprefix, index); + if (ret < 0) { + goto out; + } + + ret = dict_set_int32 (rsp_dict, key, pid); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save pid %d", pid); + goto out; + } + } + + ret = snprintf (key, sizeof (key), "%s.brick%d", + keyprefix, index); + if (ret < 0) { + goto out; + } + + ret = glusterd_get_brick_lvm_details (rsp_dict, brickinfo, + snap_volinfo->volname, + device, key); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get " + "brick LVM details"); + goto out; + } +out: + if (ret && value) { + GF_FREE (value); + } + + return ret; +} + +int +glusterd_get_single_snap_status (char **op_errstr, dict_t *rsp_dict, + char *keyprefix, glusterd_snap_t *snap) +{ + int ret = -1; + xlator_t *this = NULL; + char key[PATH_MAX] = ""; + char brickkey[PATH_MAX] = ""; + glusterd_volinfo_t *snap_volinfo = NULL; + glusterd_volinfo_t *tmp_volinfo = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + int volcount = 0; + int brickcount = 0; + + this = THIS; + GF_ASSERT (this); + + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + GF_ASSERT (keyprefix); + GF_ASSERT (snap); + + list_for_each_entry_safe (snap_volinfo, tmp_volinfo, &snap->volumes, + vol_list) { + ret = snprintf (key, sizeof (key), "%s.vol%d", keyprefix, + volcount); + if (ret < 0) { + goto out; + } + list_for_each_entry (brickinfo, &snap_volinfo->bricks, + brick_list) { + if (!glusterd_is_local_brick (this, snap_volinfo, + brickinfo)) { + brickcount++; + continue; + } + + ret = glusterd_get_single_brick_status (op_errstr, + rsp_dict, key, brickcount, + snap_volinfo, brickinfo); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Getting " + "single snap status failed"); + goto out; + } + brickcount++; + } + ret = snprintf (brickkey, sizeof (brickkey), "%s.brickcount", + key); + if (ret < 0) { + goto out; + } + + ret = dict_set_int32 (rsp_dict, brickkey, brickcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save brick count"); + goto out; + } + volcount++; + } + + ret = snprintf (key, sizeof (key), "%s.volcount", keyprefix); + if (ret < 0) { + goto out; + } + + ret = dict_set_int32 (rsp_dict, key, volcount); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save volcount"); + goto out; + } + +out: + + return ret; +} + +int +glusterd_get_each_snap_object_status (char **op_errstr, dict_t *rsp_dict, + glusterd_snap_t *snap, char *keyprefix) +{ + int ret = -1; + char key[PATH_MAX] = ""; + char *temp = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + GF_ASSERT (snap); + GF_ASSERT (keyprefix); + + /* TODO : Get all the snap volume info present in snap object, + * as of now, There will be only one snapvolinfo per snap object + */ + ret = snprintf (key, sizeof (key), "%s.snapname", keyprefix); + if (ret < 0) { + goto out; + } + + temp = gf_strdup (snap->snapname); + if (temp == NULL) { + ret = -1; + goto out; + } + ret = dict_set_dynstr (rsp_dict, key, temp); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not save " + "snap name"); + goto out; + } + + temp = NULL; + + ret = snprintf (key, sizeof (key), "%s.uuid", keyprefix); + if (ret < 0) { + goto out; + } + + temp = gf_strdup (uuid_utoa (snap->snap_id)); + if (temp == NULL) { + ret = -1; + goto out; + } + + ret = dict_set_dynstr (rsp_dict, key, temp); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not save " + "snap UUID"); + goto out; + } + + temp = NULL; + + ret = glusterd_get_single_snap_status (op_errstr, rsp_dict, keyprefix, + snap); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not get single snap status"); + goto out; + } + + ret = snprintf (key, sizeof (key), "%s.volcount", keyprefix); + if (ret < 0) { + goto out; + } + + ret = dict_set_int32 (rsp_dict, key, 1); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not save volcount"); + goto out; + } +out: + if (ret && temp) + GF_FREE (temp); + + return ret; +} + +int +glusterd_get_snap_status_of_volume (char **op_errstr, dict_t *rsp_dict, + char *volname, char *keyprefix) { + int ret = -1; + glusterd_volinfo_t *snap_volinfo = NULL; + glusterd_volinfo_t *temp_volinfo = NULL; + glusterd_volinfo_t *volinfo = NULL; + char key[PATH_MAX] = ""; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + int i = 0; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + GF_ASSERT (op_errstr); + GF_ASSERT (rsp_dict); + GF_ASSERT (volname); + GF_ASSERT (keyprefix); + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get volinfo of " + "volume %s", volname); + goto out; + } + + list_for_each_entry_safe (snap_volinfo, temp_volinfo, + &volinfo->snap_volumes, snapvol_list) { + ret = snprintf (key, sizeof (key), "status.snap%d", i); + if (ret < 0) { + goto out; + } + + ret = glusterd_get_each_snap_object_status (op_errstr, + rsp_dict, snap_volinfo->snapshot, key); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Function : " + "glusterd_get_single_snap_status failed"); + goto out; + } + i++; + } + + ret = dict_set_int32 (rsp_dict, "status.snapcount", i); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to save snapcount"); + ret = -1; + goto out; + } +out: + return ret; +} + +int +glusterd_get_all_snapshot_status (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int32_t i = 0; + int ret = -1; + char key[PATH_MAX] = ""; + glusterd_conf_t *priv = NULL; + glusterd_snap_t *snap = NULL; + glusterd_snap_t *tmp_snap = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + list_for_each_entry_safe (snap, tmp_snap, + &priv->snapshots, snap_list) { + ret = snprintf (key, sizeof (key), "status.snap%d", i); + if (ret < 0) { + goto out; + } + + ret = glusterd_get_each_snap_object_status (op_errstr, + rsp_dict, snap, key); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not get " + "the details of a snap object: %s", + snap->snapname); + goto out; + } + i++; + } + + ret = dict_set_int32 (rsp_dict, "status.snapcount", i); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Could not save snapcount"); + goto out; + } + + ret = 0; +out: + return ret; +} + + +int +glusterd_snapshot_status_commit (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + xlator_t *this = NULL; + int ret = -1; + glusterd_conf_t *conf = NULL; + char *get_buffer = NULL; + int32_t cmd = -1; + char *snapname = NULL; + glusterd_snap_t *snap = NULL; + char *volname = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (op_errstr); + + conf = this->private; + + GF_ASSERT (conf); + ret = dict_get_int32 (dict, "cmd", &cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get status cmd type"); + goto out; + } + + ret = dict_set_int32 (rsp_dict, "cmd", cmd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Could not save status cmd in rsp dictionary"); + goto out; + } + switch (cmd) { + case GF_SNAP_STATUS_TYPE_ALL: + { + ret = glusterd_get_all_snapshot_status (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "get snapshot status"); + goto out; + } + break; + } + case GF_SNAP_STATUS_TYPE_SNAP: + { + + ret = dict_get_str (dict, "snapname", &snapname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "get snap name"); + goto out; + } + + snap = glusterd_find_snap_by_name (snapname); + if (!snap) { + ret = gf_asprintf (op_errstr, "Snap (%s) " + "not found", snapname); + if (ret < 0) { + goto out; + } + ret = -1; + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "get snap volinfo"); + goto out; + } + ret = glusterd_get_each_snap_object_status (op_errstr, + rsp_dict, snap, "status.snap0"); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "get status of snap %s", get_buffer); + goto out; + } + break; + } + case GF_SNAP_STATUS_TYPE_VOL: + { + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to" + " get volume name"); + goto out; + } + + ret = glusterd_get_snap_status_of_volume (op_errstr, + rsp_dict, volname, "status.vol0"); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Function :" + " glusterd_get_snap_status_of_volume " + "failed"); + goto out; + } + } + } + ret = 0; +out: + return ret; +} + +int32_t +glusterd_snapshot_create_postvalidate (dict_t *dict, int32_t op_ret, + char **op_errstr, dict_t *rsp_dict) +{ + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + int ret = -1; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + priv = this->private; + GF_ASSERT (priv); + + if (op_ret) { + ret = glusterd_do_snap_cleanup (dict, op_errstr, rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "cleanup operation " + "failed"); + goto out; + } + } else { + ret = glusterd_snapshot_update_snaps_post_validate (dict, + op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "create snapshot"); + goto out; + } + } + + ret = 0; +out: + return ret; +} + +int32_t +glusterd_snapshot (dict_t *dict, char **op_errstr, dict_t *rsp_dict) +{ + + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + int32_t snap_command = 0; + int ret = -1; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + priv = this->private; + GF_ASSERT (priv); + + ret = dict_get_int32 (dict, "type", &snap_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case (GF_SNAP_OPTION_TYPE_CREATE): + ret = glusterd_snapshot_create_commit (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "create snapshot"); + goto out; + } + break; + + case GF_SNAP_OPTION_TYPE_CONFIG: + ret = glusterd_snapshot_config_commit (dict, op_errstr, + rsp_dict); + break; + + case GF_SNAP_OPTION_TYPE_DELETE: + ret = glusterd_snapshot_remove_commit (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "delete snapshot"); + goto out; + } + break; + + case GF_SNAP_OPTION_TYPE_RESTORE: + ret = glusterd_snapshot_restore (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Failed to " + "restore snapshot"); + goto out; + } + + break; + + case GF_SNAP_OPTION_TYPE_STATUS: + ret = glusterd_snapshot_status_commit (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "show snapshot status"); + goto out; + } + break; + + + default: + gf_log (this->name, GF_LOG_WARNING, "invalid snap command"); + goto out; + break; + } + + ret = 0; + +out: + return ret; +} + +int +glusterd_snapshot_brickop (dict_t *dict, char **op_errstr, dict_t *rsp_dict) +{ + int ret = -1; + int64_t vol_count = 0; + int64_t count = 1; + char key[1024] = {0,}; + char *volname = NULL; + int32_t snap_command = 0; + xlator_t *this = NULL; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + ret = dict_get_int32 (dict, "type", &snap_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case GF_SNAP_OPTION_TYPE_CREATE: + ret = dict_get_int64 (dict, "volcount", &vol_count); + if (ret) + goto out; + while (count <= vol_count) { + snprintf (key, 1024, "volname%"PRId64, count); + ret = dict_get_str (dict, key, &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get volname"); + goto out; + } + ret = dict_set_str (dict, "volname", volname); + if (ret) + goto out; + + ret = gd_brick_op_phase (GD_OP_SNAP, NULL, dict, + op_errstr); + if (ret) + goto out; + volname = NULL; + count++; + } + + dict_del (dict, "volname"); + ret = 0; + break; + case GF_SNAP_OPTION_TYPE_DELETE: + break; + default: + break; + } + +out: + return ret; +} + +int +glusterd_snapshot_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict) +{ + int snap_command = 0; + xlator_t *this = NULL; + int ret = -1; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + ret = dict_get_int32 (dict, "type", &snap_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case (GF_SNAP_OPTION_TYPE_CREATE): + ret = glusterd_snapshot_create_prevalidate (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot create " + "pre-validation failed"); + goto out; + } + break; + + case (GF_SNAP_OPTION_TYPE_CONFIG): + ret = glusterd_snapshot_config_prevalidate (dict, op_errstr); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot config " + "pre-validation failed"); + goto out; + } + break; + + case GF_SNAP_OPTION_TYPE_RESTORE: + ret = glusterd_snapshot_restore_prevalidate (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot restore " + "validation failed"); + goto out; + } + break; + case GF_SNAP_OPTION_TYPE_DELETE: + ret = glusterd_snapshot_remove_prevalidate (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot remove " + "validation failed"); + goto out; + } + break; + + case GF_SNAP_OPTION_TYPE_STATUS: + ret = glusterd_snapshot_status_prevalidate (dict, op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot status " + "validation failed"); + goto out; + } + break; + + default: + gf_log (this->name, GF_LOG_WARNING, "invalid snap command"); + goto out; + } + + ret = 0; +out: + return ret; +} + +int +glusterd_snapshot_postvalidate (dict_t *dict, int32_t op_ret, char **op_errstr, + dict_t *rsp_dict) +{ + int snap_command = 0; + xlator_t *this = NULL; + int ret = -1; + + this = THIS; + + GF_ASSERT (this); + GF_ASSERT (dict); + GF_ASSERT (rsp_dict); + + ret = dict_get_int32 (dict, "type", &snap_command); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case GF_SNAP_OPTION_TYPE_CREATE: + ret = glusterd_snapshot_create_postvalidate (dict, op_ret, + op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot create " + "post-validation failed"); + goto out; + } + break; + + case GF_SNAP_OPTION_TYPE_DELETE: + case GF_SNAP_OPTION_TYPE_RESTORE: + ret = glusterd_snapshot_update_snaps_post_validate (dict, + op_errstr, + rsp_dict); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "update missed snaps list"); + goto out; + } + break; + + default: + gf_log (this->name, GF_LOG_WARNING, "invalid snap command"); + goto out; + } + + ret = 0; +out: + return ret; +} + +int +glusterd_handle_snapshot_fn (rpcsvc_request_t *req) +{ + int32_t ret = 0; + dict_t *dict = NULL; + gf_cli_req cli_req = {{0},}; + glusterd_op_t cli_op = GD_OP_SNAP; + int type = 0; + glusterd_conf_t *conf = NULL; + char *host_uuid = NULL; + char err_str[2048] = {0,}; + xlator_t *this = NULL; + char *volname = NULL; + + GF_ASSERT (req); + + this = THIS; + GF_ASSERT (this); + conf = this->private; + GF_ASSERT (conf); + + ret = xdr_to_generic (req->msg[0], &cli_req, + (xdrproc_t)xdr_gf_cli_req); + if (ret < 0) { + req->rpc_err = GARBAGE_ARGS; + goto out; + } + + if (cli_req.dict.dict_len > 0) { + dict = dict_new (); + if (!dict) + goto out; + + ret = dict_unserialize (cli_req.dict.dict_val, + cli_req.dict.dict_len, + &dict); + if (ret < 0) { + gf_log (this->name, GF_LOG_ERROR, "failed to " + "unserialize req-buffer to dictionary"); + snprintf (err_str, sizeof (err_str), "Unable to decode " + "the command"); + goto out; + } + + dict->extra_stdfree = cli_req.dict.dict_val; + + host_uuid = gf_strdup (uuid_utoa(MY_UUID)); + if (host_uuid == NULL) { + snprintf (err_str, sizeof (err_str), "Failed to get " + "the uuid of local glusterd"); + ret = -1; + goto out; + } + ret = dict_set_dynstr (dict, "host-uuid", host_uuid); + if (ret) { + GF_FREE (host_uuid); + goto out; + } + + + } else { + gf_log (this->name, GF_LOG_ERROR, "request dict length is %d", + cli_req.dict.dict_len); + goto out; + } + + ret = dict_get_int32 (dict, "type", &type); + if (ret < 0) { + snprintf (err_str, sizeof (err_str), "Command type not found"); + gf_log (this->name, GF_LOG_ERROR, "%s", err_str); + goto out; + } + + switch (type) { + case GF_SNAP_OPTION_TYPE_CREATE: + ret = glusterd_handle_snapshot_create (req, cli_op, dict, + err_str, sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot create " + "failed: %s", err_str); + } + break; + case GF_SNAP_OPTION_TYPE_RESTORE: + ret = glusterd_handle_snapshot_restore (req, cli_op, dict, + err_str, sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot restore " + "failed: %s", err_str); + } + + break; + case GF_SNAP_OPTION_TYPE_INFO: + ret = glusterd_handle_snapshot_info (req, cli_op, dict, + err_str, sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot info " + "failed"); + } + break; + case GF_SNAP_OPTION_TYPE_LIST: + ret = glusterd_handle_snapshot_list (req, cli_op, dict, + err_str, sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot list " + "failed"); + } + break; + case GF_SNAP_OPTION_TYPE_CONFIG: + /* TODO : Type of lock to be taken when we are setting + * limits system wide + */ + ret = dict_get_str (dict, "volname", &volname); + if (!volname) { + ret = dict_set_int32 (dict, "hold_vol_locks", + _gf_false); + if (ret) { + gf_log ("cli", GF_LOG_ERROR, + "Unable to set hold_vol_locks value " + "as _gf_false"); + goto out; + } + + } + ret = glusterd_mgmt_v3_initiate_all_phases (req, cli_op, dict); + break; + case GF_SNAP_OPTION_TYPE_DELETE: + ret = glusterd_handle_snapshot_remove (req, cli_op, dict, + err_str, + sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot delete " + "failed: %s", err_str); + } + break; + case GF_SNAP_OPTION_TYPE_START: + case GF_SNAP_OPTION_TYPE_STOP: + case GF_SNAP_OPTION_TYPE_STATUS: + ret = glusterd_handle_snapshot_status (req, cli_op, dict, + err_str, + sizeof (err_str)); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Snapshot status " + "failed: %s", err_str); + } + break; + default: + gf_log (this->name, GF_LOG_ERROR, "Unkown snapshot request " + "type (%d)", type); + ret = -1; /* Failure */ + } + +out: + if (ret) { + if (err_str[0] == '\0') + snprintf (err_str, sizeof (err_str), + "Operation failed"); + ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, + dict, err_str); + } + + return ret; +} + +int +glusterd_handle_snapshot (rpcsvc_request_t *req) +{ + return glusterd_big_locked_handler (req, glusterd_handle_snapshot_fn); +} + +static inline void +glusterd_free_snap_op (glusterd_snap_op_t *snap_op) +{ + if (snap_op) { + if (snap_op->brick_path) + GF_FREE (snap_op->brick_path); + + GF_FREE (snap_op); + } +} + +/* Look for duplicates and accordingly update the list */ +int32_t +glusterd_update_missed_snap_entry (glusterd_missed_snap_info *missed_snapinfo, + glusterd_snap_op_t *missed_snap_op) +{ + int32_t ret = -1; + glusterd_snap_op_t *snap_opinfo = NULL; + gf_boolean_t match = _gf_false; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(missed_snapinfo); + GF_ASSERT(missed_snap_op); + + list_for_each_entry (snap_opinfo, &missed_snapinfo->snap_ops, + snap_ops_list) { + if ((!strcmp (snap_opinfo->brick_path, + missed_snap_op->brick_path)) && + (snap_opinfo->op == missed_snap_op->op)) { + /* If two entries have conflicting status + * GD_MISSED_SNAP_DONE takes precedence + */ + if ((snap_opinfo->status == GD_MISSED_SNAP_PENDING) && + (missed_snap_op->status == GD_MISSED_SNAP_DONE)) { + snap_opinfo->status = GD_MISSED_SNAP_DONE; + gf_log (this->name, GF_LOG_INFO, + "Updating missed snap status " + "for %s:%d:%s:%d as DONE", + missed_snapinfo->node_snap_info, + snap_opinfo->brick_num, + snap_opinfo->brick_path, + snap_opinfo->op); + ret = 0; + glusterd_free_snap_op (missed_snap_op); + goto out; + } + match = _gf_true; + break; + } else if ((snap_opinfo->brick_num == + missed_snap_op->brick_num) && + (snap_opinfo->op == GF_SNAP_OPTION_TYPE_CREATE) && + (missed_snap_op->op == + GF_SNAP_OPTION_TYPE_DELETE)) { + /* Optimizing create and delete entries for the same + * brick and same node + */ + gf_log (this->name, GF_LOG_INFO, + "Updating missed snap status " + "for %s:%d:%s:%d as DONE", + missed_snapinfo->node_snap_info, + snap_opinfo->brick_num, + snap_opinfo->brick_path, + snap_opinfo->op); + snap_opinfo->status = GD_MISSED_SNAP_DONE; + ret = 0; + glusterd_free_snap_op (missed_snap_op); + goto out; + } + } + + if (match == _gf_true) { + gf_log (this->name, GF_LOG_INFO, + "Duplicate entry. Not updating"); + glusterd_free_snap_op (missed_snap_op); + } else { + list_add_tail (&missed_snap_op->snap_ops_list, + &missed_snapinfo->snap_ops); + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Add new missed snap entry to the missed_snaps list. */ +int32_t +glusterd_store_missed_snaps_list (char *missed_info, int32_t brick_num, + char *brick_path, int32_t snap_op, + int32_t snap_status) +{ + int32_t ret = -1; + glusterd_missed_snap_info *missed_snapinfo = NULL; + glusterd_snap_op_t *missed_snap_op = NULL; + glusterd_conf_t *priv = NULL; + gf_boolean_t match = _gf_false; + gf_boolean_t free_missed_snap_info = _gf_false; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(missed_info); + GF_ASSERT(brick_path); + + priv = this->private; + GF_ASSERT (priv); + + /* Create the snap_op object consisting of the * + * snap id and the op */ + ret = glusterd_missed_snap_op_new (&missed_snap_op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create new missed snap object."); + ret = -1; + goto out; + } + + missed_snap_op->brick_path = gf_strdup(brick_path); + if (!missed_snap_op->brick_path) { + ret = -1; + goto out; + } + missed_snap_op->brick_num = brick_num; + missed_snap_op->op = snap_op; + missed_snap_op->status = snap_status; + + /* Look for other entries for the same node and same snap */ + list_for_each_entry (missed_snapinfo, &priv->missed_snaps_list, + missed_snaps) { + if (!strcmp (missed_snapinfo->node_snap_info, + missed_info)) { + /* Found missed snapshot info for * + * the same node and same snap */ + match = _gf_true; + break; + } + } + + if (match == _gf_false) { + /* First snap op missed for the brick */ + ret = glusterd_missed_snapinfo_new (&missed_snapinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create missed snapinfo"); + goto out; + } + free_missed_snap_info = _gf_true; + missed_snapinfo->node_snap_info = gf_strdup(missed_info); + if (!missed_snapinfo->node_snap_info) { + ret = -1; + goto out; + } + + list_add_tail (&missed_snap_op->snap_ops_list, + &missed_snapinfo->snap_ops); + list_add_tail (&missed_snapinfo->missed_snaps, + &priv->missed_snaps_list); + + ret = 0; + goto out; + } else { + ret = glusterd_update_missed_snap_entry (missed_snapinfo, + missed_snap_op); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to update existing missed snap entry."); + goto out; + } + } + +out: + if (ret) { + glusterd_free_snap_op (missed_snap_op); + + if (missed_snapinfo && + (free_missed_snap_info == _gf_true)) { + if (missed_snapinfo->node_snap_info) + GF_FREE (missed_snapinfo->node_snap_info); + + GF_FREE (missed_snapinfo); + } + } + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Add missing snap entries to the in-memory conf->missed_snap_list */ +int32_t +glusterd_add_missed_snaps_to_list (dict_t *dict, int32_t missed_snap_count) +{ + char *buf = NULL; + char *tmp = NULL; + char *save_ptr = NULL; + char *nodeid = NULL; + char *snap_uuid = NULL; + char *brick_path = NULL; + char missed_info[PATH_MAX] = ""; + char name_buf[PATH_MAX] = ""; + int32_t i = -1; + int32_t ret = -1; + int32_t brick_num = -1; + int32_t snap_op = -1; + int32_t snap_status = -1; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(dict); + + priv = this->private; + GF_ASSERT (priv); + + /* We can update the missed_snaps_list without acquiring * + * any additional locks as big lock will be held. */ + for (i = 0; i < missed_snap_count; i++) { + snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d", + i); + ret = dict_get_str (dict, name_buf, &buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch %s", name_buf); + goto out; + } + + gf_log (this->name, GF_LOG_DEBUG, "missed_snap_entry = %s", + buf); + + /* Need to make a duplicate string coz the same dictionary * + * is resent to the non-originator nodes */ + tmp = gf_strdup (buf); + if (!tmp) { + ret = -1; + goto out; + } + + /* Fetch the node-id, snap-id, brick_num, + * brick_path, snap_op and snap status + */ + nodeid = strtok_r (tmp, ":", &save_ptr); + snap_uuid = strtok_r (NULL, "=", &save_ptr); + brick_num = atoi(strtok_r (NULL, ":", &save_ptr)); + brick_path = strtok_r (NULL, ":", &save_ptr); + snap_op = atoi(strtok_r (NULL, ":", &save_ptr)); + snap_status = atoi(strtok_r (NULL, ":", &save_ptr)); + + if (!nodeid || !snap_uuid || !brick_path || + brick_num < 1 || snap_op < 1 || + snap_status < 1) { + gf_log (this->name, GF_LOG_ERROR, + "Invalid missed_snap_entry"); + ret = -1; + goto out; + } + + snprintf (missed_info, sizeof(missed_info), "%s:%s", + nodeid, snap_uuid); + + ret = glusterd_store_missed_snaps_list (missed_info, + brick_num, + brick_path, + snap_op, + snap_status); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to store missed snaps_list"); + goto out; + } + + GF_FREE (tmp); + tmp = NULL; + } + +out: + if (tmp) + GF_FREE (tmp); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-store.c b/xlators/mgmt/glusterd/src/glusterd-store.c index 2e6dd6b43..1c2ec58e8 100644 --- a/xlators/mgmt/glusterd/src/glusterd-store.c +++ b/xlators/mgmt/glusterd/src/glusterd-store.c @@ -1,5 +1,5 @@ /* - Copyright (c) 2007-2012 Red Hat, Inc. <http://www.redhat.com> + Copyright (c) 2007-2013 Red Hat, Inc. <http://www.redhat.com> This file is part of GlusterFS. This file is licensed to you under your choice of the GNU Lesser @@ -46,7 +46,7 @@ #include <dirent.h> void -glusterd_replace_slash_with_hipen (char *str) +glusterd_replace_slash_with_hyphen (char *str) { char *ptr = NULL; @@ -85,7 +85,7 @@ glusterd_store_key_vol_brick_set (glusterd_brickinfo_t *brickinfo, GF_ASSERT (len >= PATH_MAX); snprintf (key_vol_brick, len, "%s", brickinfo->path); - glusterd_replace_slash_with_hipen (key_vol_brick); + glusterd_replace_slash_with_hyphen (key_vol_brick); } static void @@ -241,6 +241,25 @@ glusterd_store_brickinfo_write (int fd, glusterd_brickinfo_t *brickinfo) if (ret) goto out; + if (strlen(brickinfo->device_path) > 0) { + snprintf (value, sizeof(value), "%s", brickinfo->device_path); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH, + value); + if (ret) + goto out; + } + + snprintf (value, sizeof(value), "%d", brickinfo->snap_status); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS, + value); + if (ret) + goto out; + + if (!brickinfo->vg[0]) + goto out; + + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_VGNAME, + brickinfo->vg); out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; @@ -297,6 +316,7 @@ glusterd_store_brickinfo (glusterd_volinfo_t *volinfo, goto out; ret = glusterd_store_perform_brick_store (brickinfo); + out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; @@ -501,14 +521,13 @@ int _storeopts (dict_t *this, char *key, data_t *value, void *data) int32_t glusterd_volume_exclude_options_write (int fd, glusterd_volinfo_t *volinfo) { - char *str = NULL; + char *str = NULL; + char buf[PATH_MAX] = {0,}; + int32_t ret = -1; GF_ASSERT (fd > 0); GF_ASSERT (volinfo); - char buf[PATH_MAX] = {0,}; - int32_t ret = -1; - snprintf (buf, sizeof (buf), "%d", volinfo->type); ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_TYPE, buf); if (ret) @@ -550,6 +569,14 @@ glusterd_volume_exclude_options_write (int fd, glusterd_volinfo_t *volinfo) if (ret) goto out; + snprintf (buf, sizeof (buf), "%s", volinfo->parent_volname); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_PARENT_VOLNAME, buf); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to store " + GLUSTERD_STORE_KEY_PARENT_VOLNAME); + goto out; + } + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_ID, uuid_utoa (volinfo->volume_id)); if (ret) @@ -571,14 +598,41 @@ glusterd_volume_exclude_options_write (int fd, glusterd_volinfo_t *volinfo) goto out; } - if (volinfo->backend == GD_VOL_BK_BD) { - snprintf (buf, sizeof (buf), "%d", volinfo->backend); - ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_BACKEND, + snprintf (buf, sizeof (buf), "%d", volinfo->op_version); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_OP_VERSION, buf); + if (ret) + goto out; + + snprintf (buf, sizeof (buf), "%d", volinfo->client_op_version); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION, + buf); + if (ret) + goto out; + if (volinfo->caps) { + snprintf (buf, sizeof (buf), "%d", volinfo->caps); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_CAPS, buf); if (ret) goto out; } + snprintf (buf, sizeof (buf), "%d", volinfo->is_volume_restored); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_IS_RESTORED, buf); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Unable to write is_volume_restored"); + goto out; + } + + snprintf (buf, sizeof (buf), "%"PRIu64, volinfo->snap_max_hard_limit); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT, + buf); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Unable to write snap-max-hard-limit"); + goto out; + } + out: if (ret) gf_log (THIS->name, GF_LOG_ERROR, "Unable to write volume " @@ -596,8 +650,7 @@ glusterd_store_voldirpath_set (glusterd_volinfo_t *volinfo, char *voldirpath, priv = THIS->private; GF_ASSERT (priv); - snprintf (voldirpath, len, "%s/%s/%s", priv->workdir, - GLUSTERD_VOLUME_DIR_PREFIX, volinfo->volname); + GLUSTERD_GET_VOLUME_DIR (voldirpath, volinfo, priv); } static int32_t @@ -611,10 +664,32 @@ glusterd_store_create_volume_dir (glusterd_volinfo_t *volinfo) glusterd_store_voldirpath_set (volinfo, voldirpath, sizeof (voldirpath)); ret = gf_store_mkdir (voldirpath); + gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret); return ret; } +static int32_t +glusterd_store_create_snap_dir (glusterd_snap_t *snap) +{ + int32_t ret = -1; + char snapdirpath[PATH_MAX] = {0,}; + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + GF_ASSERT (priv); + GF_ASSERT (snap); + + GLUSTERD_GET_SNAP_DIR (snapdirpath, snap, priv); + + ret = mkdir_p (snapdirpath, 0755, _gf_true); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to create snaps dir " + "%s", snapdirpath); + } + return ret; +} + int32_t glusterd_store_volinfo_write (int fd, glusterd_volinfo_t *volinfo) { @@ -639,6 +714,49 @@ out: return ret; } +int32_t +glusterd_store_snapinfo_write (glusterd_snap_t *snap) +{ + int32_t ret = -1; + int fd = 0; + char buf[PATH_MAX] = ""; + + GF_ASSERT (snap); + + fd = gf_store_mkstemp (snap->shandle); + if (fd <= 0) + goto out; + + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_ID, + uuid_utoa (snap->snap_id)); + if (ret) + goto out; + + snprintf (buf, sizeof (buf), "%d", snap->snap_status); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_STATUS, buf); + if (ret) + goto out; + + snprintf (buf, sizeof (buf), "%d", snap->snap_restored); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_RESTORED, buf); + if (ret) + goto out; + + if (snap->description) { + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_DESC, + snap->description); + if (ret) + goto out; + } + + snprintf (buf, sizeof (buf), "%ld", snap->time_stamp); + ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_TIMESTAMP, buf); + +out: + gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + static void glusterd_store_rbstatepath_set (glusterd_volinfo_t *volinfo, char *rbstatepath, size_t len) @@ -683,6 +801,37 @@ glusterd_store_node_state_path_set (glusterd_volinfo_t *volinfo, GLUSTERD_NODE_STATE_FILE); } +static void +glusterd_store_missed_snaps_list_path_set (char *missed_snaps_list, + size_t len) +{ + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + GF_ASSERT (priv); + GF_ASSERT (missed_snaps_list); + GF_ASSERT (len <= PATH_MAX); + + snprintf (missed_snaps_list, len, "%s/snaps/" + GLUSTERD_MISSED_SNAPS_LIST_FILE, priv->workdir); +} + +static void +glusterd_store_snapfpath_set (glusterd_snap_t *snap, char *snap_fpath, + size_t len) +{ + glusterd_conf_t *priv = NULL; + + priv = THIS->private; + GF_ASSERT (priv); + GF_ASSERT (snap); + GF_ASSERT (snap_fpath); + GF_ASSERT (len <= PATH_MAX); + + snprintf (snap_fpath, len, "%s/snaps/%s/%s", priv->workdir, + snap->snapname, GLUSTERD_SNAP_INFO_FILE); +} + int32_t glusterd_store_create_rbstate_shandle_on_absence (glusterd_volinfo_t *volinfo) { @@ -727,6 +876,42 @@ glusterd_store_create_nodestate_sh_on_absence (glusterd_volinfo_t *volinfo) return ret; } +static int32_t +glusterd_store_create_missed_snaps_list_shandle_on_absence () +{ + char missed_snaps_list[PATH_MAX] = ""; + int32_t ret = -1; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + priv = this->private; + GF_ASSERT (priv); + + glusterd_store_missed_snaps_list_path_set (missed_snaps_list, + sizeof(missed_snaps_list)); + + ret = gf_store_handle_create_on_absence + (&priv->missed_snaps_list_shandle, + missed_snaps_list); + return ret; +} + +int32_t +glusterd_store_create_snap_shandle_on_absence (glusterd_snap_t *snap) +{ + char snapfpath[PATH_MAX] = {0}; + int32_t ret = 0; + + GF_ASSERT (snap); + + glusterd_store_snapfpath_set (snap, snapfpath, sizeof (snapfpath)); + ret = gf_store_handle_create_on_absence (&snap->shandle, snapfpath); + return ret; +} + int32_t glusterd_store_brickinfos (glusterd_volinfo_t *volinfo, int vol_fd) { @@ -974,6 +1159,14 @@ glusterd_store_volume_cleanup_tmp (glusterd_volinfo_t *volinfo) gf_store_unlink_tmppath (volinfo->node_state_shandle); } +void +glusterd_store_snap_cleanup_tmp (glusterd_snap_t *snap) +{ + GF_ASSERT (snap); + + gf_store_unlink_tmppath (snap->shandle); +} + int32_t glusterd_store_brickinfos_atomic_update (glusterd_volinfo_t *volinfo) { @@ -1025,6 +1218,60 @@ out: } int32_t +glusterd_store_snap_atomic_update (glusterd_snap_t *snap) +{ + int ret = -1; + GF_ASSERT (snap); + + ret = gf_store_rename_tmppath (snap->shandle); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, "Couldn't rename " + "temporary file(s): Reason %s", strerror (errno)); + + return ret; +} + +int32_t +glusterd_store_snap (glusterd_snap_t *snap) +{ + int32_t ret = -1; + + GF_ASSERT (snap); + + ret = glusterd_store_create_snap_dir (snap); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to create snap dir"); + goto out; + } + + ret = glusterd_store_create_snap_shandle_on_absence (snap); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to create snap info " + "file"); + goto out; + } + + ret = glusterd_store_snapinfo_write (snap); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to write snap info"); + goto out; + } + + ret = glusterd_store_snap_atomic_update (snap); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR,"Failed to do automic update"); + goto out; + } + +out: + if (ret) + glusterd_store_snap_cleanup_tmp (snap); + + gf_log (THIS->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t glusterd_store_volinfo (glusterd_volinfo_t *volinfo, glusterd_volinfo_ver_ac_t ac) { int32_t ret = -1; @@ -1081,7 +1328,6 @@ out: return ret; } - int32_t glusterd_store_delete_volume (glusterd_volinfo_t *volinfo) { @@ -1108,8 +1354,8 @@ glusterd_store_delete_volume (glusterd_volinfo_t *volinfo) GLUSTERD_GET_VOLUME_DIR (pathname, volinfo, priv); snprintf (delete_path, sizeof (delete_path), - "%s/"GLUSTERD_TRASH"/%s.deleted", priv->workdir, - uuid_utoa (volinfo->volume_id)); + "%s/"GLUSTERD_TRASH"/%s.deleted", priv->workdir, + uuid_utoa (volinfo->volume_id)); snprintf (trashdir, sizeof (trashdir), "%s/"GLUSTERD_TRASH, priv->workdir); @@ -1201,6 +1447,116 @@ out: return ret; } +/*TODO: cleanup the duplicate code and implement a generic function for + * deleting snap/volume depending on the parameter flag */ +int32_t +glusterd_store_delete_snap (glusterd_snap_t *snap) +{ + char pathname[PATH_MAX] = {0,}; + int32_t ret = 0; + glusterd_conf_t *priv = NULL; + DIR *dir = NULL; + struct dirent *entry = NULL; + char path[PATH_MAX] = {0,}; + char delete_path[PATH_MAX] = {0,}; + char trashdir[PATH_MAX] = {0,}; + struct stat st = {0, }; + xlator_t *this = NULL; + gf_boolean_t rename_fail = _gf_false; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + + GF_ASSERT (snap); + GLUSTERD_GET_SNAP_DIR (pathname, snap, priv); + + snprintf (delete_path, sizeof (delete_path), + "%s/"GLUSTERD_TRASH"/snap-%s.deleted", priv->workdir, + uuid_utoa (snap->snap_id)); + + snprintf (trashdir, sizeof (trashdir), "%s/"GLUSTERD_TRASH, + priv->workdir); + + ret = mkdir (trashdir, 0777); + if (ret && errno != EEXIST) { + gf_log (this->name, GF_LOG_ERROR, "Failed to create trash " + "directory, reason : %s", strerror (errno)); + ret = -1; + goto out; + } + + ret = rename (pathname, delete_path); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to rename snap " + "directory %s to %s", snap->snapname, delete_path); + rename_fail = _gf_true; + goto out; + } + + dir = opendir (delete_path); + if (!dir) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to open directory %s." + " Reason : %s", delete_path, strerror (errno)); + ret = 0; + goto out; + } + + glusterd_for_each_entry (entry, dir); + while (entry) { + snprintf (path, PATH_MAX, "%s/%s", delete_path, entry->d_name); + ret = stat (path, &st); + if (ret == -1) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to stat " + "entry %s : %s", path, strerror (errno)); + goto stat_failed; + } + + if (S_ISDIR (st.st_mode)) + ret = rmdir (path); + else + ret = unlink (path); + + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, " Failed to remove " + "%s. Reason : %s", path, strerror (errno)); + } + + gf_log (this->name, GF_LOG_DEBUG, "%s %s", + ret ? "Failed to remove":"Removed", + entry->d_name); +stat_failed: + memset (path, 0, sizeof(path)); + glusterd_for_each_entry (entry, dir); + } + + ret = closedir (dir); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to close dir %s. " + "Reason : %s",delete_path, strerror (errno)); + } + + ret = rmdir (delete_path); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to rmdir: %s,err: %s", + delete_path, strerror (errno)); + } + ret = rmdir (trashdir); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to rmdir: %s, Reason:" + " %s", trashdir, strerror (errno)); + } + +out: + if (snap->shandle) { + gf_store_handle_destroy (snap->shandle); + snap->shandle = NULL; + } + ret = (rename_fail == _gf_true) ? -1: 0; + + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} int glusterd_store_global_info (xlator_t *this) @@ -1211,6 +1567,7 @@ glusterd_store_global_info (xlator_t *this) char path[PATH_MAX] = {0,}; gf_store_handle_t *handle = NULL; char *uuid_str = NULL; + char buf[256] = {0, }; conf = this->private; @@ -1263,6 +1620,24 @@ glusterd_store_global_info (xlator_t *this) goto out; } + snprintf (buf, sizeof (buf), "%"PRIu64, conf->snap_max_hard_limit); + ret = gf_store_save_value (handle->fd, + GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT, buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Storing snap-max-hard-limit failed ret = %d", ret); + goto out; + } + + snprintf (buf, sizeof (buf), "%"PRIu64, conf->snap_max_soft_limit); + ret = gf_store_save_value (handle->fd, + GLUSTERD_STORE_KEY_SNAP_MAX_SOFT_LIMIT, buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Storing snap-max-soft-limit failed ret = %d", ret); + goto out; + } + ret = gf_store_rename_tmppath (handle); out: if (ret && (handle->fd > 0)) @@ -1334,15 +1709,94 @@ out: return ret; } +int +glusterd_retrieve_sys_snap_max_limit (xlator_t *this, uint64_t *limit, + char *key) +{ + char *limit_str = NULL; + glusterd_conf_t *priv = NULL; + int ret = -1; + uint64_t tmp_limit = 0; + char *tmp = NULL; + char path[PATH_MAX] = {0,}; + gf_store_handle_t *handle = NULL; + + GF_ASSERT (this); + priv = this->private; + + GF_ASSERT (priv); + GF_ASSERT (limit); + GF_ASSERT (key); + + if (!priv->handle) { + snprintf (path, PATH_MAX, "%s/%s", priv->workdir, + GLUSTERD_INFO_FILE); + ret = gf_store_handle_retrieve (path, &handle); + + if (ret) { + gf_log ("", GF_LOG_DEBUG, "Unable to get store " + "handle!"); + goto out; + } + + priv->handle = handle; + } + + ret = gf_store_retrieve_value (priv->handle, + key, + &limit_str); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "No previous %s present", key); + goto out; + } + + tmp_limit = strtoul (limit_str, &tmp, 10); + if ((tmp_limit <= 0) || (tmp && strlen (tmp) > 1)) { + gf_log (this->name, GF_LOG_WARNING, "invalid version number"); + goto out; + } + + *limit = tmp_limit; + + ret = 0; +out: + if (limit_str) + GF_FREE (limit_str); + + return ret; +} static int glusterd_restore_op_version (xlator_t *this) { - glusterd_conf_t *conf = NULL; - int ret = 0; - int op_version = 0; + glusterd_conf_t *conf = NULL; + int ret = 0; + int op_version = 0; conf = this->private; + ret = glusterd_retrieve_sys_snap_max_limit (this, + &conf->snap_max_hard_limit, + GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Unable to retrieve system snap-max-hard-limit, " + "setting it to default value(%d)", + GLUSTERD_SNAPS_MAX_HARD_LIMIT); + conf->snap_max_hard_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT; + } + + ret = glusterd_retrieve_sys_snap_max_limit (this, + &conf->snap_max_soft_limit, + GLUSTERD_STORE_KEY_SNAP_MAX_SOFT_LIMIT); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Unable to retrieve system snap-max-soft-limit, " + "setting it to default value(%d)", + GLUSTERD_SNAPS_DEF_SOFT_LIMIT_PERCENT); + conf->snap_max_soft_limit = GLUSTERD_SNAPS_DEF_SOFT_LIMIT_PERCENT; + } + ret = glusterd_retrieve_op_version (this, &op_version); if (!ret) { if ((op_version < GD_OP_VERSION_MIN) || @@ -1431,7 +1885,6 @@ out: int32_t glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) { - int32_t ret = 0; glusterd_brickinfo_t *brickinfo = NULL; gf_store_iter_t *iter = NULL; @@ -1452,10 +1905,9 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) priv = THIS->private; - GLUSTERD_GET_BRICK_DIR (brickdir, volinfo, priv) + GLUSTERD_GET_BRICK_DIR (brickdir, volinfo, priv); ret = gf_store_iter_new (volinfo->shandle, &tmpiter); - if (ret) goto out; @@ -1502,7 +1954,7 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) strlen (GLUSTERD_STORE_KEY_BRICK_PORT))) { gf_string2int (value, &brickinfo->port); - if (brickinfo->port < GF_IANA_PRIV_PORTS_START){ + if (brickinfo->port < priv->base_port) { /* This is required to adhere to the IANA standards */ brickinfo->port = 0; @@ -1518,8 +1970,7 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) strlen (GLUSTERD_STORE_KEY_BRICK_RDMA_PORT))) { gf_string2int (value, &brickinfo->rdma_port); - if (brickinfo->rdma_port < - GF_IANA_PRIV_PORTS_START){ + if (brickinfo->rdma_port < priv->base_port) { /* This is required to adhere to the IANA standards */ brickinfo->rdma_port = 0; @@ -1536,6 +1987,18 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED, strlen (GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED))) { gf_string2int (value, &brickinfo->decommissioned); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH, + strlen (GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH))) { + strncpy (brickinfo->device_path, value, + sizeof (brickinfo->device_path)); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS, + strlen (GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS))) { + gf_string2int (value, &brickinfo->snap_status); + } else if (!strncmp (key, + GLUSTERD_STORE_KEY_BRICK_VGNAME, + strlen (GLUSTERD_STORE_KEY_BRICK_VGNAME))) { + strncpy (brickinfo->vg, value, + sizeof (brickinfo->vg)); } else { gf_log ("", GF_LOG_ERROR, "Unknown key: %s", key); @@ -1572,10 +2035,9 @@ out: int32_t -glusterd_store_retrieve_rbstate (char *volname) +glusterd_store_retrieve_rbstate (glusterd_volinfo_t *volinfo) { int32_t ret = -1; - glusterd_volinfo_t *volinfo = NULL; gf_store_iter_t *iter = NULL; char *key = NULL; char *value = NULL; @@ -1583,15 +2045,13 @@ glusterd_store_retrieve_rbstate (char *volname) glusterd_conf_t *priv = NULL; char path[PATH_MAX] = {0,}; gf_store_op_errno_t op_errno = GD_STORE_SUCCESS; + xlator_t *this = NULL; - priv = THIS->private; - - ret = glusterd_volinfo_find (volname, &volinfo); - if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "Couldn't get" - "volinfo for %s.", volname); - goto out; - } + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (volinfo); GLUSTERD_GET_VOLUME_DIR(volpath, volinfo, priv); snprintf (path, sizeof (path), "%s/%s", volpath, @@ -1668,16 +2128,15 @@ glusterd_store_retrieve_rbstate (char *volname) goto out; out: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); return ret; } int32_t -glusterd_store_retrieve_node_state (char *volname) +glusterd_store_retrieve_node_state (glusterd_volinfo_t *volinfo) { int32_t ret = -1; - glusterd_volinfo_t *volinfo = NULL; gf_store_iter_t *iter = NULL; char *key = NULL; char *value = NULL; @@ -1685,15 +2144,13 @@ glusterd_store_retrieve_node_state (char *volname) glusterd_conf_t *priv = NULL; char path[PATH_MAX] = {0,}; gf_store_op_errno_t op_errno = GD_STORE_SUCCESS; + xlator_t *this = NULL; - priv = THIS->private; - - ret = glusterd_volinfo_find (volname, &volinfo); - if (ret) { - gf_log (THIS->name, GF_LOG_ERROR, "Couldn't get" - "volinfo for %s.", volname); - goto out; - } + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (volinfo); GLUSTERD_GET_VOLUME_DIR(volpath, volinfo, priv); snprintf (path, sizeof (path), "%s/%s", volpath, @@ -1745,53 +2202,58 @@ glusterd_store_retrieve_node_state (char *volname) goto out; out: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); return ret; } -int32_t -glusterd_store_retrieve_volume (char *volname) -{ - int32_t ret = -1; - glusterd_volinfo_t *volinfo = NULL; - gf_store_iter_t *iter = NULL; - char *key = NULL; - char *value = NULL; - char volpath[PATH_MAX] = {0,}; - glusterd_conf_t *priv = NULL; - char path[PATH_MAX] = {0,}; - int exists = 0; - gf_store_op_errno_t op_errno = GD_STORE_SUCCESS; - - ret = glusterd_volinfo_new (&volinfo); - if (ret) - goto out; +int +glusterd_store_update_volinfo (glusterd_volinfo_t *volinfo) +{ + int ret = -1; + int exists = 0; + char *key = NULL; + char *value = NULL; + char volpath[PATH_MAX] = {0,}; + char path[PATH_MAX] = {0,}; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + gf_store_iter_t *iter = NULL; + gf_store_op_errno_t op_errno = GD_STORE_SUCCESS; - strncpy (volinfo->volname, volname, GLUSTERD_MAX_VOLUME_NAME); + this = THIS; + GF_ASSERT (this); + conf = THIS->private; + GF_ASSERT (volinfo); - priv = THIS->private; + GLUSTERD_GET_VOLUME_DIR(volpath, volinfo, conf); - GLUSTERD_GET_VOLUME_DIR(volpath, volinfo, priv); snprintf (path, sizeof (path), "%s/%s", volpath, GLUSTERD_VOLUME_INFO_FILE); ret = gf_store_handle_retrieve (path, &volinfo->shandle); - - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "volinfo handle is NULL"); goto out; + } ret = gf_store_iter_new (volinfo->shandle, &iter); - - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get new store " + "iter"); goto out; + } ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); - if (ret) + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get next store " + "iter"); goto out; + } while (!ret) { + gf_log ("", GF_LOG_DEBUG, "key = %s value = %s", key, value); if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_TYPE, strlen (GLUSTERD_STORE_KEY_VOL_TYPE))) { volinfo->type = atoi (value); @@ -1851,9 +2313,24 @@ glusterd_store_retrieve_volume (char *volname) gf_log ("", GF_LOG_DEBUG, "Parsed as "GEOREP" " " slave:key=%s,value:%s", key, value); - } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_BACKEND, - strlen (GLUSTERD_STORE_KEY_VOL_BACKEND))) { - volinfo->backend = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_OP_VERSION, + strlen (GLUSTERD_STORE_KEY_VOL_OP_VERSION))) { + volinfo->op_version = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION, + strlen (GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION))) { + volinfo->client_op_version = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_CAPS, + strlen (GLUSTERD_STORE_KEY_VOL_CAPS))) { + volinfo->caps = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT, + strlen (GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT))) { + volinfo->snap_max_hard_limit = (uint64_t) atoll (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_IS_RESTORED, + strlen (GLUSTERD_STORE_KEY_VOL_IS_RESTORED))) { + volinfo->is_volume_restored = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_PARENT_VOLNAME, + strlen (GLUSTERD_STORE_KEY_PARENT_VOLNAME))) { + strncpy (volinfo->parent_volname, value, sizeof(volinfo->parent_volname) - 1); } else { if (is_key_glusterd_hooks_friendly (key)) { @@ -1932,16 +2409,59 @@ glusterd_store_retrieve_volume (char *volname) volinfo->subvol_count = (volinfo->brick_count / volinfo->dist_leaf_count); + /* Only calculate volume op-versions if they are not found */ + if (!volinfo->op_version && !volinfo->client_op_version) + gd_update_volume_op_versions (volinfo); } if (op_errno != GD_STORE_EOF) goto out; ret = gf_store_iter_destroy (iter); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to destroy store " + "iter"); + goto out; + } + + ret = 0; +out: + return ret; +} +glusterd_volinfo_t* +glusterd_store_retrieve_volume (char *volname, glusterd_snap_t *snap) +{ + int32_t ret = -1; + glusterd_volinfo_t *volinfo = NULL; + glusterd_volinfo_t *origin_volinfo = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (volname); + + ret = glusterd_volinfo_new (&volinfo); if (ret) goto out; + priv = THIS->private; + + strncpy (volinfo->volname, volname, GLUSTERD_MAX_VOLUME_NAME); + volinfo->snapshot = snap; + if (snap) + volinfo->is_snap_volume = _gf_true; + + ret = glusterd_store_update_volinfo (volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to update volinfo " + "for %s volume", volname); + goto out; + } + ret = glusterd_store_retrieve_bricks (volinfo); if (ret) goto out; @@ -1950,14 +2470,28 @@ glusterd_store_retrieve_volume (char *volname) if (ret) goto out; - gd_update_volume_op_versions (volinfo); - - list_add_tail (&volinfo->vol_list, &priv->volumes); - + if (!snap) { + list_add_tail (&volinfo->vol_list, &priv->volumes); + } else { + ret = glusterd_volinfo_find (volinfo->parent_volname, + &origin_volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Parent volinfo " + "not found for %s volume", volname); + goto out; + } + glusterd_list_add_snapvol (origin_volinfo, volinfo); + } out: - gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + if (ret) { + if (volinfo) + glusterd_volinfo_delete (volinfo); + volinfo = NULL; + } - return ret; + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); + + return volinfo; } inline void @@ -2057,9 +2591,9 @@ out: } int32_t -glusterd_store_retrieve_volumes (xlator_t *this) +glusterd_store_retrieve_volumes (xlator_t *this, glusterd_snap_t *snap) { - int32_t ret = 0; + int32_t ret = -1; char path[PATH_MAX] = {0,}; glusterd_conf_t *priv = NULL; DIR *dir = NULL; @@ -2071,51 +2605,414 @@ glusterd_store_retrieve_volumes (xlator_t *this) GF_ASSERT (priv); - snprintf (path, PATH_MAX, "%s/%s", priv->workdir, - GLUSTERD_VOLUME_DIR_PREFIX); + if (snap) + snprintf (path, PATH_MAX, "%s/snaps/%s", priv->workdir, + snap->snapname); + else + snprintf (path, PATH_MAX, "%s/%s", priv->workdir, + GLUSTERD_VOLUME_DIR_PREFIX); dir = opendir (path); if (!dir) { gf_log ("", GF_LOG_ERROR, "Unable to open dir %s", path); - ret = -1; goto out; } glusterd_for_each_entry (entry, dir); while (entry) { - ret = glusterd_store_retrieve_volume (entry->d_name); - if (ret) { + if ( entry->d_type != DT_DIR ) + goto next; + + volinfo = glusterd_store_retrieve_volume (entry->d_name, snap); + if (!volinfo) { gf_log ("", GF_LOG_ERROR, "Unable to restore " "volume: %s", entry->d_name); + ret = -1; goto out; } - ret = glusterd_store_retrieve_rbstate (entry->d_name); + ret = glusterd_store_retrieve_rbstate (volinfo); if (ret) { /* Backward compatibility */ gf_log ("", GF_LOG_INFO, "Creating a new rbstate " "for volume: %s.", entry->d_name); - ret = glusterd_volinfo_find (entry->d_name, &volinfo); ret = glusterd_store_create_rbstate_shandle_on_absence (volinfo); ret = glusterd_store_perform_rbstate_store (volinfo); } - ret = glusterd_store_retrieve_node_state (entry->d_name); + ret = glusterd_store_retrieve_node_state (volinfo); if (ret) { /* Backward compatibility */ gf_log ("", GF_LOG_INFO, "Creating a new node_state " "for volume: %s.", entry->d_name); - ret = glusterd_volinfo_find (entry->d_name, &volinfo); - ret = glusterd_store_create_nodestate_sh_on_absence (volinfo); ret = glusterd_store_perform_node_state_store (volinfo); } + +next: + glusterd_for_each_entry (entry, dir); + } + + ret = 0; +out: + if (dir) + closedir (dir); + gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret); + + return ret; +} + +int32_t +glusterd_resolve_snap_bricks (xlator_t *this, glusterd_snap_t *snap) +{ + int32_t ret = -1; + glusterd_volinfo_t *volinfo = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + + GF_ASSERT (this); + GF_VALIDATE_OR_GOTO (this->name, snap, out); + + list_for_each_entry (volinfo, &snap->volumes, vol_list) { + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + ret = glusterd_resolve_brick (brickinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "resolve brick failed in restore"); + goto out; + } + } + } + + ret = 0; + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); + + return ret; +} + +int +glusterd_store_update_snap (glusterd_snap_t *snap) +{ + int ret = -1; + char *key = NULL; + char *value = NULL; + char snappath[PATH_MAX] = {0,}; + char path[PATH_MAX] = {0,}; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + gf_store_iter_t *iter = NULL; + gf_store_op_errno_t op_errno = GD_STORE_SUCCESS; + + this = THIS; + conf = this->private; + GF_ASSERT (snap); + + GLUSTERD_GET_SNAP_DIR (snappath, snap, conf); + + snprintf (path, sizeof (path), "%s/%s", snappath, + GLUSTERD_SNAP_INFO_FILE); + + ret = gf_store_handle_retrieve (path, &snap->shandle); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "snap handle is NULL"); + goto out; + } + + ret = gf_store_iter_new (snap->shandle, &iter); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get new store " + "iter"); + goto out; + } + + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get next store " + "iter"); + goto out; + } + + while (!ret) { + gf_log (this->name, GF_LOG_DEBUG, "key = %s value = %s", + key, value); + + if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_ID, + strlen (GLUSTERD_STORE_KEY_SNAP_ID))) { + ret = uuid_parse (value, snap->snap_id); + if (ret) + gf_log (this->name, GF_LOG_WARNING, + "Failed to parse uuid"); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_RESTORED, + strlen (GLUSTERD_STORE_KEY_SNAP_RESTORED))) { + snap->snap_restored = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_STATUS, + strlen (GLUSTERD_STORE_KEY_SNAP_STATUS))) { + snap->snap_status = atoi (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_DESC, + strlen (GLUSTERD_STORE_KEY_SNAP_DESC))) { + snap->description = gf_strdup (value); + } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_TIMESTAMP, + strlen (GLUSTERD_STORE_KEY_SNAP_TIMESTAMP))) { + snap->time_stamp = atoi (value); + } + + GF_FREE (key); + GF_FREE (value); + key = NULL; + value = NULL; + + ret = gf_store_iter_get_next (iter, &key, &value, &op_errno); + } + + if (op_errno != GD_STORE_EOF) + goto out; + + ret = gf_store_iter_destroy (iter); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to destroy store " + "iter"); + } + +out: + return ret; +} + +int32_t +glusterd_store_retrieve_snap (char *snapname) +{ + int32_t ret = -1; + dict_t *dict = NULL; + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (snapname); + + dict = dict_new(); + if (!dict) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create dict"); + ret = -1; + goto out; + } + + snap = glusterd_new_snap_object (); + if (!snap) { + gf_log (this->name, GF_LOG_ERROR, "Failed to create " + " snap object"); + goto out; + } + + strncpy (snap->snapname, snapname, strlen(snapname)); + ret = glusterd_store_update_snap (snap); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to update snapshot " + "for %s snap", snapname); + goto out; + } + + ret = glusterd_store_retrieve_volumes (this, snap); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to retrieve " + "snap volumes for snap %s", snapname); + goto out; + } + + /* Unlike bricks of normal volumes which are resolved at the end of + the glusterd restore, the bricks belonging to the snap volumes of + each snap should be resolved as part of snapshot restore itself. + Because if the snapshot has to be removed, then resolving bricks + helps glusterd in understanding what all bricks have its own uuid + and killing those bricks. + */ + ret = glusterd_resolve_snap_bricks (this, snap); + if (ret) + gf_log (this->name, GF_LOG_WARNING, "resolving the snap bricks" + " failed (snap: %s)", snap?snap->snapname:""); + + /* When the snapshot command from cli is received, the on disk and + in memory structures for the snapshot are created (with the status) + being marked as GD_SNAP_STATUS_INIT. Once the backend snapshot is + taken, the status is changed to GD_SNAP_STATUS_IN_USE. If glusterd + dies after taking the backend snapshot, but before updating the + status, then when glusterd comes up, it should treat that snapshot + as a failed snapshot and clean it up. + */ + if (snap->snap_status != GD_SNAP_STATUS_IN_USE) { + ret = glusterd_snap_remove (dict, snap, _gf_true, _gf_true); + if (ret) + gf_log (this->name, GF_LOG_WARNING, "failed to remove" + " the snapshot %s", snap->snapname); + goto out; + } + + /* TODO: list_add_order can do 'N-square' comparisions and + is not efficient. Find a better solution to store the snap + in order */ + list_add_order (&snap->snap_list, &priv->snapshots, + glusterd_compare_snap_time); + +out: + if (dict) + dict_unref (dict); + + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); + return ret; +} + +/* Read the missed_snap_list and update the in-memory structs */ +int32_t +glusterd_store_retrieve_missed_snaps_list (xlator_t *this) +{ + char buf[PATH_MAX] = ""; + char path[PATH_MAX] = ""; + char *missed_node_info = NULL; + char *brick_path = NULL; + char *value = NULL; + char *save_ptr = NULL; + FILE *fp = NULL; + int32_t brick_num = -1; + int32_t snap_op = -1; + int32_t snap_status = -1; + int32_t ret = -1; + glusterd_conf_t *priv = NULL; + gf_store_op_errno_t store_errno = GD_STORE_SUCCESS; + + GF_ASSERT (this); + priv = this->private; + GF_ASSERT (priv); + + /* Get the path of the missed_snap_list */ + glusterd_store_missed_snaps_list_path_set (path, sizeof(path)); + + fp = fopen (path, "r"); + if (!fp) { + /* If errno is ENOENT then there are no missed snaps yet */ + if (errno != ENOENT) { + ret = -1; + gf_log (this->name, GF_LOG_ERROR, + "Failed to open %s. " + "Error: %s", path, strerror(errno)); + } else { + gf_log (this->name, GF_LOG_INFO, + "No missed snaps list."); + ret = 0; + } + goto out; + } + + do { + ret = gf_store_read_and_tokenize (fp, buf, + &missed_node_info, &value, + &store_errno); + if (ret) { + if (store_errno == GD_STORE_EOF) { + gf_log (this->name, + GF_LOG_DEBUG, + "EOF for missed_snap_list"); + ret = 0; + break; + } + gf_log (this->name, GF_LOG_ERROR, + "Failed to fetch data from " + "missed_snaps_list. Error: %s", + gf_store_strerror (store_errno)); + goto out; + } + + /* Fetch the brick_num, brick_path, snap_op and snap status */ + brick_num = atoi(strtok_r (value, ":", &save_ptr)); + brick_path = strtok_r (NULL, ":", &save_ptr); + snap_op = atoi(strtok_r (NULL, ":", &save_ptr)); + snap_status = atoi(strtok_r (NULL, ":", &save_ptr)); + + if (!missed_node_info || !brick_path || + brick_num < 1 || snap_op < 1 || + snap_status < 1) { + gf_log (this->name, GF_LOG_ERROR, + "Invalid missed_snap_entry"); + ret = -1; + goto out; + } + + ret = glusterd_store_missed_snaps_list (missed_node_info, + brick_num, + brick_path, + snap_op, + snap_status); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to store missed snaps_list"); + goto out; + } + + } while (store_errno == GD_STORE_SUCCESS); + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret); + return ret; +} + +int32_t +glusterd_store_retrieve_snaps (xlator_t *this) +{ + int32_t ret = 0; + char path[PATH_MAX] = {0,}; + glusterd_conf_t *priv = NULL; + DIR *dir = NULL; + struct dirent *entry = NULL; + + GF_ASSERT (this); + priv = this->private; + + GF_ASSERT (priv); + + snprintf (path, PATH_MAX, "%s/snaps", priv->workdir); + + dir = opendir (path); + + if (!dir) { + /* If snaps dir doesn't exists ignore the error for + backward compatibility */ + if (errno != ENOENT) { + ret = -1; + gf_log ("", GF_LOG_ERROR, "Unable to open dir %s", path); + } + goto out; + } + + glusterd_for_each_entry (entry, dir); + + while (entry) { + if (entry->d_type == DT_DIR) { + ret = glusterd_store_retrieve_snap (entry->d_name); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to restore snapshot: %s", + entry->d_name); + goto out; + } + } + glusterd_for_each_entry (entry, dir); } + /* Retrieve missed_snaps_list */ + ret = glusterd_store_retrieve_missed_snaps_list (this); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Failed to retrieve missed_snaps_list"); + goto out; + } + out: if (dir) closedir (dir); @@ -2124,6 +3021,126 @@ out: return ret; } +/* Writes all the contents of conf->missed_snap_list */ +int32_t +glusterd_store_write_missed_snapinfo (int32_t fd) +{ + char value[PATH_MAX] = ""; + int32_t ret = -1; + glusterd_conf_t *priv = NULL; + glusterd_missed_snap_info *missed_snapinfo = NULL; + glusterd_snap_op_t *snap_opinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + + priv = this->private; + GF_ASSERT (priv); + + /* Write the missed_snap_entry */ + list_for_each_entry (missed_snapinfo, &priv->missed_snaps_list, + missed_snaps) { + list_for_each_entry (snap_opinfo, + &missed_snapinfo->snap_ops, + snap_ops_list) { + snprintf (value, sizeof(value), "%d:%s:%d:%d", + snap_opinfo->brick_num, + snap_opinfo->brick_path, + snap_opinfo->op, snap_opinfo->status); + ret = gf_store_save_value + (fd, + missed_snapinfo->node_snap_info, + value); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to write missed snapinfo"); + goto out; + } + } + } + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +/* Adds the missed snap entries to the in-memory conf->missed_snap_list * + * and writes them to disk */ +int32_t +glusterd_store_update_missed_snaps (dict_t *dict, int32_t missed_snap_count) +{ + int32_t fd = -1; + int32_t ret = -1; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT(this); + GF_ASSERT(dict); + + priv = this->private; + GF_ASSERT (priv); + + if (missed_snap_count < 1) { + gf_log (this->name, GF_LOG_DEBUG, "No missed snaps"); + ret = 0; + goto out; + } + + ret = glusterd_store_create_missed_snaps_list_shandle_on_absence (); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to obtain " + "missed_snaps_list store handle."); + goto out; + } + + fd = gf_store_mkstemp (priv->missed_snaps_list_shandle); + if (fd <= 0) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to create tmp file"); + ret = -1; + goto out; + } + + ret = glusterd_add_missed_snaps_to_list (dict, missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add missed snaps to list"); + goto out; + } + + ret = glusterd_store_write_missed_snapinfo (fd); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to write missed snaps to disk"); + goto out; + } + + ret = gf_store_rename_tmppath (priv->missed_snaps_list_shandle); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to rename the tmp file"); + goto out; + } +out: + if (ret && (fd > 0)) { + ret = gf_store_unlink_tmppath (priv->missed_snaps_list_shandle); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to unlink the tmp file"); + } + ret = -1; + } + + if (fd > 0) + close (fd); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + int32_t glusterd_store_delete_peerinfo (glusterd_peerinfo_t *peerinfo) { @@ -2525,7 +3542,11 @@ glusterd_restore () goto out; } - ret = glusterd_store_retrieve_volumes (this); + ret = glusterd_store_retrieve_volumes (this, NULL); + if (ret) + goto out; + + ret = glusterd_store_retrieve_snaps (this); if (ret) goto out; diff --git a/xlators/mgmt/glusterd/src/glusterd-store.h b/xlators/mgmt/glusterd/src/glusterd-store.h index e4f9e4a0a..1b5cebc0c 100644 --- a/xlators/mgmt/glusterd/src/glusterd-store.h +++ b/xlators/mgmt/glusterd/src/glusterd-store.h @@ -35,38 +35,55 @@ typedef enum glusterd_store_ver_ac_{ } glusterd_volinfo_ver_ac_t; -#define GLUSTERD_STORE_UUID_KEY "UUID" - -#define GLUSTERD_STORE_KEY_VOL_TYPE "type" -#define GLUSTERD_STORE_KEY_VOL_COUNT "count" -#define GLUSTERD_STORE_KEY_VOL_STATUS "status" -#define GLUSTERD_STORE_KEY_VOL_PORT "port" -#define GLUSTERD_STORE_KEY_VOL_SUB_COUNT "sub_count" -#define GLUSTERD_STORE_KEY_VOL_STRIPE_CNT "stripe_count" -#define GLUSTERD_STORE_KEY_VOL_REPLICA_CNT "replica_count" -#define GLUSTERD_STORE_KEY_VOL_BRICK "brick" -#define GLUSTERD_STORE_KEY_VOL_VERSION "version" -#define GLUSTERD_STORE_KEY_VOL_TRANSPORT "transport-type" -#define GLUSTERD_STORE_KEY_VOL_ID "volume-id" -#define GLUSTERD_STORE_KEY_RB_STATUS "rb_status" -#define GLUSTERD_STORE_KEY_RB_SRC_BRICK "rb_src" -#define GLUSTERD_STORE_KEY_RB_DST_BRICK "rb_dst" -#define GLUSTERD_STORE_KEY_RB_DST_PORT "rb_port" -#define GLUSTERD_STORE_KEY_VOL_DEFRAG "rebalance_status" -#define GLUSTERD_STORE_KEY_DEFRAG_OP "rebalance_op" -#define GLUSTERD_STORE_KEY_USERNAME "username" -#define GLUSTERD_STORE_KEY_PASSWORD "password" - -#define GLUSTERD_STORE_KEY_BRICK_HOSTNAME "hostname" -#define GLUSTERD_STORE_KEY_BRICK_PATH "path" -#define GLUSTERD_STORE_KEY_BRICK_PORT "listen-port" -#define GLUSTERD_STORE_KEY_BRICK_RDMA_PORT "rdma.listen-port" +#define GLUSTERD_STORE_UUID_KEY "UUID" + +#define GLUSTERD_STORE_KEY_VOL_TYPE "type" +#define GLUSTERD_STORE_KEY_VOL_COUNT "count" +#define GLUSTERD_STORE_KEY_VOL_STATUS "status" +#define GLUSTERD_STORE_KEY_VOL_PORT "port" +#define GLUSTERD_STORE_KEY_VOL_SUB_COUNT "sub_count" +#define GLUSTERD_STORE_KEY_VOL_STRIPE_CNT "stripe_count" +#define GLUSTERD_STORE_KEY_VOL_REPLICA_CNT "replica_count" +#define GLUSTERD_STORE_KEY_VOL_BRICK "brick" +#define GLUSTERD_STORE_KEY_VOL_VERSION "version" +#define GLUSTERD_STORE_KEY_VOL_TRANSPORT "transport-type" +#define GLUSTERD_STORE_KEY_VOL_ID "volume-id" +#define GLUSTERD_STORE_KEY_VOL_IS_RESTORED "is-volume-restored" +#define GLUSTERD_STORE_KEY_RB_STATUS "rb_status" +#define GLUSTERD_STORE_KEY_RB_SRC_BRICK "rb_src" +#define GLUSTERD_STORE_KEY_RB_DST_BRICK "rb_dst" +#define GLUSTERD_STORE_KEY_RB_DST_PORT "rb_port" +#define GLUSTERD_STORE_KEY_VOL_DEFRAG "rebalance_status" +#define GLUSTERD_STORE_KEY_DEFRAG_OP "rebalance_op" +#define GLUSTERD_STORE_KEY_USERNAME "username" +#define GLUSTERD_STORE_KEY_PASSWORD "password" +#define GLUSTERD_STORE_KEY_PARENT_VOLNAME "parent_volname" +#define GLUSTERD_STORE_KEY_VOL_OP_VERSION "op-version" +#define GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION "client-op-version" + +#define GLUSTERD_STORE_KEY_SNAP_NAME "name" +#define GLUSTERD_STORE_KEY_SNAP_ID "snap-id" +#define GLUSTERD_STORE_KEY_SNAP_DESC "desc" +#define GLUSTERD_STORE_KEY_SNAP_TIMESTAMP "time-stamp" +#define GLUSTERD_STORE_KEY_SNAP_STATUS "status" +#define GLUSTERD_STORE_KEY_SNAP_RESTORED "snap-restored" +#define GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT "snap-max-hard-limit" +#define GLUSTERD_STORE_KEY_SNAP_MAX_SOFT_LIMIT "snap-max-soft-limit" + +#define GLUSTERD_STORE_KEY_BRICK_HOSTNAME "hostname" +#define GLUSTERD_STORE_KEY_BRICK_PATH "path" +#define GLUSTERD_STORE_KEY_BRICK_PORT "listen-port" +#define GLUSTERD_STORE_KEY_BRICK_RDMA_PORT "rdma.listen-port" #define GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED "decommissioned" +#define GLUSTERD_STORE_KEY_BRICK_VGNAME "vg" +#define GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH "device_path" +#define GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS "snap-status" -#define GLUSTERD_STORE_KEY_PEER_UUID "uuid" -#define GLUSTERD_STORE_KEY_PEER_HOSTNAME "hostname" -#define GLUSTERD_STORE_KEY_PEER_STATE "state" -#define GLUSTERD_STORE_KEY_VOL_BACKEND "backend" +#define GLUSTERD_STORE_KEY_PEER_UUID "uuid" +#define GLUSTERD_STORE_KEY_PEER_HOSTNAME "hostname" +#define GLUSTERD_STORE_KEY_PEER_STATE "state" + +#define GLUSTERD_STORE_KEY_VOL_CAPS "caps" #define glusterd_for_each_entry(entry, dir) \ do {\ @@ -89,6 +106,9 @@ int32_t glusterd_store_delete_volume (glusterd_volinfo_t *volinfo); int32_t +glusterd_store_delete_snap (glusterd_snap_t *snap); + +int32_t glusterd_retrieve_uuid (); int32_t @@ -123,5 +143,22 @@ int32_t glusterd_store_retrieve_options (xlator_t *this); int32_t +glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo); + +int32_t glusterd_store_options (xlator_t *this, dict_t *opts); + +void +glusterd_replace_slash_with_hyphen (char *str); + +int32_t +glusterd_store_perform_volume_store (glusterd_volinfo_t *volinfo); + +int32_t +glusterd_store_snap (glusterd_snap_t *snap); + +int32_t +glusterd_store_update_missed_snaps (dict_t *dict, + int32_t missed_snap_count); + #endif diff --git a/xlators/mgmt/glusterd/src/glusterd-syncop.c b/xlators/mgmt/glusterd/src/glusterd-syncop.c index a694cae84..438df8266 100644 --- a/xlators/mgmt/glusterd/src/glusterd-syncop.c +++ b/xlators/mgmt/glusterd/src/glusterd-syncop.c @@ -17,8 +17,11 @@ #include "glusterd.h" #include "glusterd-op-sm.h" #include "glusterd-utils.h" +#include "glusterd-locks.h" -static inline void +extern glusterd_op_info_t opinfo; + +void gd_synctask_barrier_wait (struct syncargs *args, int count) { glusterd_conf_t *conf = THIS->private; @@ -31,9 +34,9 @@ gd_synctask_barrier_wait (struct syncargs *args, int count) } static void -gd_collate_errors (struct syncargs *args, int op_ret, int op_errno, - char *op_errstr, int op_code, - glusterd_peerinfo_t *peerinfo, u_char *uuid) +gd_mgmt_v3_collate_errors (struct syncargs *args, int op_ret, int op_errno, + char *op_errstr, int op_code, + glusterd_peerinfo_t *peerinfo, u_char *uuid) { char err_str[PATH_MAX] = "Please check log file for details."; char op_err[PATH_MAX] = ""; @@ -56,13 +59,67 @@ gd_collate_errors (struct syncargs *args, int op_ret, int op_errno, } switch (op_code){ - case GLUSTERD_MGMT_CLUSTER_LOCK : + case GLUSTERD_MGMT_V3_LOCK: { len = snprintf (op_err, sizeof(op_err) - 1, - "Locking failed on %s. %s", - peer_str, err_str); + "Locking failed " + "on %s. %s", peer_str, err_str); + break; + } + case GLUSTERD_MGMT_V3_UNLOCK: + { + len = snprintf (op_err, sizeof(op_err) - 1, + "Unlocking failed " + "on %s. %s", peer_str, err_str); break; } + } + op_err[len] = '\0'; + + if (args->errstr) { + len = snprintf (err_str, sizeof(err_str) - 1, + "%s\n%s", args->errstr, + op_err); + GF_FREE (args->errstr); + args->errstr = NULL; + } else + len = snprintf (err_str, sizeof(err_str) - 1, + "%s", op_err); + err_str[len] = '\0'; + + gf_log ("", GF_LOG_ERROR, "%s", op_err); + args->errstr = gf_strdup (err_str); + } + + return; +} + +static void +gd_collate_errors (struct syncargs *args, int op_ret, int op_errno, + char *op_errstr, int op_code, + glusterd_peerinfo_t *peerinfo, u_char *uuid) +{ + char err_str[PATH_MAX] = "Please check log file for details."; + char op_err[PATH_MAX] = ""; + int len = -1; + char *peer_str = NULL; + + if (op_ret) { + args->op_ret = op_ret; + args->op_errno = op_errno; + + if (peerinfo) + peer_str = peerinfo->hostname; + else + peer_str = uuid_utoa (uuid); + + if (op_errstr && strcmp (op_errstr, "")) { + len = snprintf (err_str, sizeof(err_str) - 1, + "Error: %s", op_errstr); + err_str[len] = '\0'; + } + + switch (op_code){ case GLUSTERD_MGMT_CLUSTER_UNLOCK : { len = snprintf (op_err, sizeof(op_err) - 1, @@ -105,7 +162,7 @@ gd_collate_errors (struct syncargs *args, int op_ret, int op_errno, return; } -static void +void gd_syncargs_init (struct syncargs *args, dict_t *op_ctx) { args->dict = op_ctx; @@ -207,8 +264,9 @@ out: /* Defined in glusterd-rpc-ops.c */ extern struct rpc_clnt_program gd_mgmt_prog; extern struct rpc_clnt_program gd_brick_prog; +extern struct rpc_clnt_program gd_mgmt_v3_prog; -static int +int glusterd_syncop_aggr_rsp_dict (glusterd_op_t op, dict_t *aggr, dict_t *rsp) { int ret = 0; @@ -275,6 +333,12 @@ glusterd_syncop_aggr_rsp_dict (glusterd_op_t op, dict_t *aggr, dict_t *rsp) goto out; break; + case GD_OP_SNAP: + ret = glusterd_snap_use_rsp_dict (aggr, rsp); + if (ret) + goto out; + break; + default: break; } @@ -353,6 +417,183 @@ gd_syncop_mgmt_lock (glusterd_peerinfo_t *peerinfo, struct syncargs *args, } int32_t +gd_syncop_mgmt_v3_lock_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_lock_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int op_ret = -1; + int op_errno = -1; + + GF_ASSERT(req); + GF_ASSERT(iov); + GF_ASSERT(myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_LOCK, + peerinfo, rsp.uuid); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_syncop_mgmt_v3_lock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_syncop_mgmt_v3_lock_cbk_fn); +} + +int +gd_syncop_mgmt_v3_lock (glusterd_op_t op, dict_t *op_ctx, + glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid, uuid_t txn_id) +{ + int ret = -1; + gd1_mgmt_v3_lock_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + + GF_ASSERT(op_ctx); + GF_ASSERT(peerinfo); + GF_ASSERT(args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + uuid_copy (req.txn_id, txn_id); + req.op = op; + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_LOCK, + gd_syncop_mgmt_v3_lock_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_lock_req); + synclock_lock (&conf->big_lock); +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +gd_syncop_mgmt_v3_unlock_cbk_fn (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + int ret = -1; + struct syncargs *args = NULL; + glusterd_peerinfo_t *peerinfo = NULL; + gd1_mgmt_v3_unlock_rsp rsp = {{0},}; + call_frame_t *frame = NULL; + int op_ret = -1; + int op_errno = -1; + + GF_ASSERT(req); + GF_ASSERT(iov); + GF_ASSERT(myframe); + + frame = myframe; + args = frame->local; + peerinfo = frame->cookie; + frame->local = NULL; + frame->cookie = NULL; + + if (-1 == req->rpc_status) { + op_errno = ENOTCONN; + goto out; + } + + ret = xdr_to_generic (*iov, &rsp, + (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp); + if (ret < 0) + goto out; + + uuid_copy (args->uuid, rsp.uuid); + + /* Set peer as locked, so we unlock only the locked peers */ + if (rsp.op_ret == 0) + peerinfo->locked = _gf_true; + op_ret = rsp.op_ret; + op_errno = rsp.op_errno; +out: + gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL, + GLUSTERD_MGMT_V3_UNLOCK, + peerinfo, rsp.uuid); + STACK_DESTROY (frame->root); + synctask_barrier_wake(args); + return 0; +} + +int32_t +gd_syncop_mgmt_v3_unlock_cbk (struct rpc_req *req, struct iovec *iov, + int count, void *myframe) +{ + return glusterd_big_locked_cbk (req, iov, count, myframe, + gd_syncop_mgmt_v3_unlock_cbk_fn); +} + +int +gd_syncop_mgmt_v3_unlock (dict_t *op_ctx, glusterd_peerinfo_t *peerinfo, + struct syncargs *args, uuid_t my_uuid, + uuid_t recv_uuid, uuid_t txn_id) +{ + int ret = -1; + gd1_mgmt_v3_unlock_req req = {{0},}; + glusterd_conf_t *conf = THIS->private; + + GF_ASSERT(op_ctx); + GF_ASSERT(peerinfo); + GF_ASSERT(args); + + ret = dict_allocate_and_serialize (op_ctx, + &req.dict.dict_val, + &req.dict.dict_len); + if (ret) + goto out; + + uuid_copy (req.uuid, my_uuid); + uuid_copy (req.txn_id, txn_id); + synclock_unlock (&conf->big_lock); + ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo, + &gd_mgmt_v3_prog, + GLUSTERD_MGMT_V3_UNLOCK, + gd_syncop_mgmt_v3_unlock_cbk, + (xdrproc_t) xdr_gd1_mgmt_v3_unlock_req); + synclock_lock (&conf->big_lock); +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t _gd_syncop_mgmt_unlock_cbk (struct rpc_req *req, struct iovec *iov, int count, void *myframe) { @@ -641,10 +882,12 @@ gd_syncop_mgmt_brick_op (struct rpc_clnt *rpc, glusterd_pending_node_t *pnode, GD_SYNCOP (rpc, (&args), NULL, gd_syncop_brick_op_cbk, req, &gd_brick_prog, req->op, xdr_gd1_mgmt_brick_op_req); - if (args.errstr && errstr) - *errstr = args.errstr; - else - GF_FREE (args.errstr); + if (args.errstr) { + if ((strlen(args.errstr) > 0) && errstr) + *errstr = args.errstr; + else + GF_FREE (args.errstr); + } if (GD_OP_STATUS_VOLUME == op) { ret = dict_set_int32 (args.dict, "index", pnode->index); @@ -814,8 +1057,8 @@ gd_build_peers_list (struct list_head *peers, struct list_head *xact_peers, } int -gd_lock_op_phase (struct list_head *peers, glusterd_op_t op, dict_t *op_ctx, - char **op_errstr, int npeers) +gd_lock_op_phase (glusterd_conf_t *conf, glusterd_op_t op, dict_t *op_ctx, + char **op_errstr, int npeers, uuid_t txn_id) { int ret = -1; int peer_cnt = 0; @@ -823,6 +1066,9 @@ gd_lock_op_phase (struct list_head *peers, glusterd_op_t op, dict_t *op_ctx, xlator_t *this = NULL; glusterd_peerinfo_t *peerinfo = NULL; struct syncargs args = {0}; + struct list_head *peers = NULL; + + peers = &conf->xaction_peers; if (!npeers) { ret = 0; @@ -833,22 +1079,38 @@ gd_lock_op_phase (struct list_head *peers, glusterd_op_t op, dict_t *op_ctx, synctask_barrier_init((&args)); peer_cnt = 0; list_for_each_entry (peerinfo, peers, op_peers_list) { - /* Reset lock status */ - peerinfo->locked = _gf_false; - gd_syncop_mgmt_lock (peerinfo, &args, MY_UUID, peer_uuid); + if (conf->op_version < 3) { + /* Reset lock status */ + peerinfo->locked = _gf_false; + gd_syncop_mgmt_lock (peerinfo, &args, + MY_UUID, peer_uuid); + } else + gd_syncop_mgmt_v3_lock (op, op_ctx, peerinfo, &args, + MY_UUID, peer_uuid, txn_id); peer_cnt++; } gd_synctask_barrier_wait((&args), peer_cnt); - ret = args.op_ret; - if (ret) { - gf_asprintf (op_errstr, "Another transaction could be " - "in progress. Please try again after " - "sometime."); - gf_log (this->name, GF_LOG_ERROR, "Failed to acquire lock"); - goto out; + + if (args.op_ret) { + if (args.errstr) + *op_errstr = gf_strdup (args.errstr); + else { + ret = gf_asprintf (op_errstr, "Another transaction could be " + "in progress. Please try again after " + "sometime."); + if (ret == -1) + *op_errstr = NULL; + + gf_log (this->name, GF_LOG_ERROR, + "Failed to acquire lock"); + + } } - ret = 0; + ret = args.op_ret; + + gf_log (this->name, GF_LOG_DEBUG, "Sent lock op req for 'Volume %s' " + "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret); out: return ret; } @@ -1016,9 +1278,10 @@ out: } int -gd_unlock_op_phase (struct list_head *peers, glusterd_op_t op, int op_ret, +gd_unlock_op_phase (glusterd_conf_t *conf, glusterd_op_t op, int op_ret, rpcsvc_request_t *req, dict_t *op_ctx, char *op_errstr, - int npeers) + int npeers, char *volname, gf_boolean_t is_acquired, + uuid_t txn_id) { glusterd_peerinfo_t *peerinfo = NULL; glusterd_peerinfo_t *tmp = NULL; @@ -1027,23 +1290,34 @@ gd_unlock_op_phase (struct list_head *peers, glusterd_op_t op, int op_ret, int ret = -1; xlator_t *this = NULL; struct syncargs args = {0}; + struct list_head *peers = NULL; + + peers = &conf->xaction_peers; if (!npeers) { ret = 0; goto out; } + /* If the lock has not been held during this + * transaction, do not send unlock requests */ + if (!is_acquired) + goto out; + this = THIS; synctask_barrier_init((&args)); peer_cnt = 0; list_for_each_entry_safe (peerinfo, tmp, peers, op_peers_list) { - /* Only unlock peers that were locked */ - if (peerinfo->locked) { - gd_syncop_mgmt_unlock (peerinfo, &args, MY_UUID, - tmp_uuid); - peer_cnt++; - } - + if (conf->op_version < 3) { + /* Only unlock peers that were locked */ + if (peerinfo->locked) + gd_syncop_mgmt_unlock (peerinfo, &args, + MY_UUID, tmp_uuid); + } else + gd_syncop_mgmt_v3_unlock (op_ctx, peerinfo, + &args, MY_UUID, + tmp_uuid, txn_id); + peer_cnt++; list_del_init (&peerinfo->op_peers_list); } gd_synctask_barrier_wait((&args), peer_cnt); @@ -1056,7 +1330,20 @@ gd_unlock_op_phase (struct list_head *peers, glusterd_op_t op, int op_ret, out: glusterd_op_send_cli_response (op, op_ret, 0, req, op_ctx, op_errstr); glusterd_op_clear_op (op); - glusterd_unlock (MY_UUID); + if (is_acquired) { + /* Based on the op-version, we release * + * the cluster or mgmt_v3 lock */ + if (conf->op_version < 3) + glusterd_unlock (MY_UUID); + else { + ret = glusterd_mgmt_v3_unlock (volname, MY_UUID, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to release lock for %s", + volname); + } + } return 0; } @@ -1073,7 +1360,8 @@ gd_get_brick_count (struct list_head *bricks) } int -gd_brick_op_phase (glusterd_op_t op, dict_t *op_ctx, dict_t *req_dict, char **op_errstr) +gd_brick_op_phase (glusterd_op_t op, dict_t *op_ctx, dict_t *req_dict, + char **op_errstr) { glusterd_pending_node_t *pending_node = NULL; struct list_head selected = {0,}; @@ -1145,20 +1433,63 @@ out: void gd_sync_task_begin (dict_t *op_ctx, rpcsvc_request_t * req) { - int ret = -1; - int npeers = 0; - dict_t *req_dict = NULL; - glusterd_conf_t *conf = NULL; - glusterd_op_t op = 0; - int32_t tmp_op = 0; - char *op_errstr = NULL; - xlator_t *this = NULL; + int ret = -1; + int npeers = 0; + dict_t *req_dict = NULL; + glusterd_conf_t *conf = NULL; + glusterd_op_t op = 0; + int32_t tmp_op = 0; + char *op_errstr = NULL; + char *tmp = NULL; + char *volname = NULL; + xlator_t *this = NULL; + gf_boolean_t is_acquired = _gf_false; + uuid_t *txn_id = NULL; + uuid_t *originator_uuid = NULL; + glusterd_op_info_t txn_opinfo; this = THIS; GF_ASSERT (this); conf = this->private; GF_ASSERT (conf); + /* Generate a transaction-id for this operation and + * save it in the dict */ + txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + if (!txn_id) { + ret = -1; + goto out; + } + + uuid_generate (*txn_id); + + ret = dict_set_bin (op_ctx, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set transaction id."); + goto out; + } else + gf_log (this->name, GF_LOG_DEBUG, + "Transaction_id = %s", uuid_utoa (*txn_id)); + + /* Save the MY_UUID as the originator_uuid */ + originator_uuid = GF_CALLOC (1, sizeof(uuid_t), + gf_common_mt_uuid_t); + if (!originator_uuid) { + ret = -1; + goto out; + } + + uuid_copy (*originator_uuid, MY_UUID); + ret = dict_set_bin (op_ctx, "originator_uuid", + originator_uuid, sizeof (uuid_t)); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set originator_uuid."); + goto out; + } + ret = dict_get_int32 (op_ctx, GD_SYNC_OPCODE_KEY, &tmp_op); if (ret) { gf_log (this->name, GF_LOG_ERROR, "Failed to get volume " @@ -1167,24 +1498,77 @@ gd_sync_task_begin (dict_t *op_ctx, rpcsvc_request_t * req) } op = tmp_op; - ret = glusterd_lock (MY_UUID); - if (ret) { - gf_log (this->name, GF_LOG_ERROR, "Unable to acquire lock"); - gf_asprintf (&op_errstr, "Another transaction is in progress. " - "Please try again after sometime."); - goto out; + + /* Based on the op_version, acquire a cluster or mgmt_v3 lock */ + if (conf->op_version < 3) { + ret = glusterd_lock (MY_UUID); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire lock"); + gf_asprintf (&op_errstr, + "Another transaction is in progress. " + "Please try again after sometime."); + goto out; + } + } else { + + /* If no volname is given as a part of the command, locks will + * not be held */ + ret = dict_get_str (op_ctx, "volname", &tmp); + if (ret) { + gf_log ("", GF_LOG_DEBUG, "Failed to get volume " + "name"); + goto local_locking_done; + } else { + /* Use a copy of volname, as cli response will be + * sent before the unlock, and the volname in the + * dict, might be removed */ + volname = gf_strdup (tmp); + if (!volname) + goto out; + } + + ret = glusterd_mgmt_v3_lock (volname, MY_UUID, "vol"); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire lock for %s", volname); + gf_asprintf (&op_errstr, + "Another transaction is in progress " + "for %s. Please try again after sometime.", + volname); + goto out; + } } - /* storing op globally to access in synctask code paths - * This is still acceptable, as we are performing this under - * the 'cluster' lock*/ - glusterd_op_set_op (op); - INIT_LIST_HEAD (&conf->xaction_peers); - npeers = gd_build_peers_list (&conf->peers, &conf->xaction_peers, op); + is_acquired = _gf_true; + +local_locking_done: - ret = gd_lock_op_phase (&conf->xaction_peers, op, op_ctx, &op_errstr, npeers); + /* Save opinfo for this transaction with the transaction id */ + glusterd_txn_opinfo_init (&txn_opinfo, NULL, &op, NULL, NULL); + ret = glusterd_set_txn_opinfo (txn_id, &txn_opinfo); if (ret) - goto out; + gf_log (this->name, GF_LOG_ERROR, + "Unable to set transaction's opinfo"); + + opinfo = txn_opinfo; + + INIT_LIST_HEAD (&conf->xaction_peers); + + /* Make 'volume status tasks' command a local operation. + * This is accomplished by setting npeers to 0. + */ + if (!glusterd_is_status_tasks_op (op, op_ctx)) + npeers = gd_build_peers_list (&conf->peers, + &conf->xaction_peers, op); + + /* If no volname is given as a part of the command, locks will + * not be held */ + if (volname) { + ret = gd_lock_op_phase (conf, op, op_ctx, &op_errstr, npeers, *txn_id); + if (ret) + goto out; + } ret = glusterd_op_build_payload (&req_dict, &op_errstr, op_ctx); if (ret) { @@ -1211,8 +1595,17 @@ gd_sync_task_begin (dict_t *op_ctx, rpcsvc_request_t * req) ret = 0; out: - (void) gd_unlock_op_phase (&conf->xaction_peers, op, ret, req, - op_ctx, op_errstr, npeers); + (void) gd_unlock_op_phase (conf, op, ret, req, op_ctx, op_errstr, + npeers, volname, is_acquired, *txn_id); + + /* Clearing the transaction opinfo */ + ret = glusterd_clear_txn_opinfo (txn_id); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to clear transaction's opinfo"); + + if (volname) + GF_FREE (volname); if (req_dict) dict_unref (req_dict); diff --git a/xlators/mgmt/glusterd/src/glusterd-syncop.h b/xlators/mgmt/glusterd/src/glusterd-syncop.h index 7e4d47f9a..e83ea2f4c 100644 --- a/xlators/mgmt/glusterd/src/glusterd-syncop.h +++ b/xlators/mgmt/glusterd/src/glusterd-syncop.h @@ -12,6 +12,7 @@ #include "syncop.h" #include "glusterd-sm.h" +#include "glusterd.h" #define GD_SYNC_OPCODE_KEY "sync-mgmt-operation" @@ -51,4 +52,20 @@ int gd_syncop_mgmt_stage_op (struct rpc_clnt *rpc, struct syncargs *arg, int gd_syncop_mgmt_commit_op (struct rpc_clnt *rpc, struct syncargs *arg, uuid_t my_uuid, uuid_t recv_uuid, int op, dict_t *dict_out, dict_t *op_ctx); + +void +gd_synctask_barrier_wait (struct syncargs *args, int count); + +int +gd_build_peers_list (struct list_head *peers, struct list_head *xact_peers, + glusterd_op_t op); +int +gd_brick_op_phase (glusterd_op_t op, dict_t *op_ctx, dict_t *req_dict, + char **op_errstr); + +int +glusterd_syncop_aggr_rsp_dict (glusterd_op_t op, dict_t *aggr, dict_t *rsp); + +void +gd_syncargs_init (struct syncargs *args, dict_t *op_ctx); #endif /* __RPC_SYNCOP_H */ diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c index bde5b9b52..e8ae05851 100644 --- a/xlators/mgmt/glusterd/src/glusterd-utils.c +++ b/xlators/mgmt/glusterd/src/glusterd-utils.c @@ -35,6 +35,8 @@ #include "glusterd-store.h" #include "glusterd-volgen.h" #include "glusterd-pmap.h" +#include "glusterfs-acl.h" +#include "glusterd-locks.h" #include "xdr-generic.h" #include <sys/resource.h> @@ -48,6 +50,11 @@ #include <unistd.h> #include <fnmatch.h> #include <sys/statvfs.h> +#include <ifaddrs.h> +#ifdef HAVE_BD_XLATOR +#include <lvm2app.h> +#endif + #ifdef GF_LINUX_HOST_OS #include <mntent.h> @@ -68,9 +75,6 @@ #define NLMV4_VERSION 4 #define NLMV1_VERSION 1 -#define ACL_PROGRAM 100227 -#define ACLV3_VERSION 3 - #define CEILING_POS(X) (((X)-(int)(X)) > 0 ? (int)((X)+1) : (int)(X)) static glusterd_lock_t lock; @@ -432,8 +436,11 @@ glusterd_volinfo_new (glusterd_volinfo_t **volinfo) if (!new_volinfo) goto out; + LOCK_INIT (&new_volinfo->lock); INIT_LIST_HEAD (&new_volinfo->vol_list); + INIT_LIST_HEAD (&new_volinfo->snapvol_list); INIT_LIST_HEAD (&new_volinfo->bricks); + INIT_LIST_HEAD (&new_volinfo->snap_volumes); new_volinfo->dict = dict_new (); if (!new_volinfo->dict) { @@ -449,6 +456,10 @@ glusterd_volinfo_new (glusterd_volinfo_t **volinfo) goto out; } + snprintf (new_volinfo->parent_volname, GLUSTERD_MAX_VOLUME_NAME, "N/A"); + + new_volinfo->snap_max_hard_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT; + new_volinfo->xl = THIS; *volinfo = new_volinfo; @@ -460,6 +471,219 @@ out: return ret; } +/* This function will create a new volinfo and then + * dup the entries from volinfo to the new_volinfo. + * + * @param volinfo volinfo which will be duplicated + * @param dup_volinfo new volinfo which will be created + * @param set_userauth if this true then auth info is also set + * + * @return 0 on success else -1 + */ +int32_t +glusterd_volinfo_dup (glusterd_volinfo_t *volinfo, + glusterd_volinfo_t **dup_volinfo, + gf_boolean_t set_userauth) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_volinfo_t *new_volinfo = NULL; + + this = THIS; + GF_ASSERT (this); + GF_VALIDATE_OR_GOTO (this->name, volinfo, out); + GF_VALIDATE_OR_GOTO (this->name, dup_volinfo, out); + + ret = glusterd_volinfo_new (&new_volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "not able to create the " + "duplicate volinfo for the volume %s", + volinfo->volname); + goto out; + } + + new_volinfo->type = volinfo->type; + new_volinfo->replica_count = volinfo->replica_count; + new_volinfo->stripe_count = volinfo->stripe_count; + new_volinfo->dist_leaf_count = volinfo->dist_leaf_count; + new_volinfo->sub_count = volinfo->sub_count; + new_volinfo->transport_type = volinfo->transport_type; + new_volinfo->nfs_transport_type = volinfo->nfs_transport_type; + new_volinfo->brick_count = volinfo->brick_count; + + dict_copy (volinfo->dict, new_volinfo->dict); + gd_update_volume_op_versions (new_volinfo); + + if (set_userauth) { + glusterd_auth_set_username (new_volinfo, + volinfo->auth.username); + glusterd_auth_set_password (new_volinfo, + volinfo->auth.password); + } + + *dup_volinfo = new_volinfo; + ret = 0; +out: + if (ret && (NULL != new_volinfo)) { + (void) glusterd_volinfo_delete (new_volinfo); + } + return ret; +} + +/* This function will duplicate brickinfo + * + * @param brickinfo Source brickinfo + * @param dup_brickinfo Destination brickinfo + * + * @return 0 on success else -1 + */ +int32_t +glusterd_brickinfo_dup (glusterd_brickinfo_t *brickinfo, + glusterd_brickinfo_t *dup_brickinfo) +{ + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + GF_VALIDATE_OR_GOTO (this->name, brickinfo, out); + GF_VALIDATE_OR_GOTO (this->name, dup_brickinfo, out); + + strcpy (dup_brickinfo->hostname, brickinfo->hostname); + strcpy (dup_brickinfo->path, brickinfo->path); + strcpy (dup_brickinfo->device_path, brickinfo->device_path); + ret = gf_canonicalize_path (dup_brickinfo->path); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to canonicalize " + "brick path"); + goto out; + } + uuid_copy (dup_brickinfo->uuid, brickinfo->uuid); + + dup_brickinfo->port = brickinfo->port; + dup_brickinfo->rdma_port = brickinfo->rdma_port; + if (NULL != brickinfo->logfile) { + dup_brickinfo->logfile = gf_strdup (brickinfo->logfile); + if (NULL == dup_brickinfo->logfile) { + ret = -1; + goto out; + } + } + dup_brickinfo->status = brickinfo->status; + dup_brickinfo->snap_status = brickinfo->snap_status; +out: + return ret; +} + +/* This function will copy snap volinfo to the new + * passed volinfo and regenerate backend store files + * for the restored snap. + * + * @param new_volinfo new volinfo + * @param snap_volinfo volinfo of snap volume + * + * @return 0 on success and -1 on failure + * + * TODO: Duplicate all members of volinfo, e.g. geo-rep sync slaves + */ +int32_t +glusterd_snap_volinfo_restore (dict_t *rsp_dict, + glusterd_volinfo_t *new_volinfo, + glusterd_volinfo_t *snap_volinfo) +{ + int32_t brick_count = -1; + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_brickinfo_t *new_brickinfo = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + + GF_VALIDATE_OR_GOTO (this->name, new_volinfo, out); + GF_VALIDATE_OR_GOTO (this->name, snap_volinfo, out); + + brick_count = 0; + list_for_each_entry (brickinfo, &snap_volinfo->bricks, brick_list) { + ret = glusterd_brickinfo_new (&new_brickinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to create " + "new brickinfo"); + goto out; + } + + /* Duplicate brickinfo */ + ret = glusterd_brickinfo_dup (brickinfo, new_brickinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to dup " + "brickinfo"); + goto out; + } + + /* If the brick is not of this peer, or snapshot is missed * + * for the brick do not replace the xattr for it */ + if ((!uuid_compare (brickinfo->uuid, MY_UUID)) && + (brickinfo->snap_status != -1)) { + /* We need to replace the volume id of all the bricks + * to the volume id of the origin volume. new_volinfo + * has the origin volume's volume id*/ + ret = sys_lsetxattr (new_brickinfo->path, + GF_XATTR_VOL_ID_KEY, + new_volinfo->volume_id, + sizeof (new_volinfo->volume_id), + XATTR_REPLACE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to " + "set extended attribute %s on %s. " + "Reason: %s, snap: %s", + GF_XATTR_VOL_ID_KEY, + new_brickinfo->path, strerror (errno), + new_volinfo->volname); + goto out; + } + } + + /* If a snapshot is pending for this brick then + * restore should also be pending + */ + if (brickinfo->snap_status == -1) { + /* Adding missed delete to the dict */ + ret = glusterd_add_missed_snaps_to_dict + (rsp_dict, + snap_volinfo->volname, + brickinfo, + brick_count + 1, + GF_SNAP_OPTION_TYPE_RESTORE); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add missed snapshot info " + "for %s:%s in the rsp_dict", + brickinfo->hostname, + brickinfo->path); + goto out; + } + } + + list_add_tail (&new_brickinfo->brick_list, + &new_volinfo->bricks); + /* ownership of new_brickinfo is passed to new_volinfo */ + new_brickinfo = NULL; + brick_count++; + } + + /* Regenerate all volfiles */ + ret = glusterd_create_volfiles_and_notify_services (new_volinfo); + +out: + if (ret && (NULL != new_brickinfo)) { + (void) glusterd_brickinfo_delete (new_brickinfo); + } + + return ret; +} + void glusterd_auth_cleanup (glusterd_volinfo_t *volinfo) { @@ -552,6 +776,7 @@ glusterd_volinfo_delete (glusterd_volinfo_t *volinfo) GF_ASSERT (volinfo); list_del_init (&volinfo->vol_list); + list_del_init (&volinfo->snapvol_list); ret = glusterd_volume_brickinfos_delete (volinfo); if (ret) @@ -572,7 +797,6 @@ out: return ret; } - int32_t glusterd_brickinfo_new (glusterd_brickinfo_t **brickinfo) { @@ -624,6 +848,7 @@ glusterd_brickinfo_new_from_brick (char *brick, char *path = NULL; char *tmp_host = NULL; char *tmp_path = NULL; + char *vg = NULL; GF_ASSERT (brick); GF_ASSERT (brickinfo); @@ -642,6 +867,17 @@ glusterd_brickinfo_new_from_brick (char *brick, if (ret) goto out; +#ifdef HAVE_BD_XLATOR + vg = strchr (path, '?'); + /* ? is used as a delimiter for vg */ + if (vg) { + strncpy (new_brickinfo->vg, vg + 1, PATH_MAX - 1); + *vg = '\0'; + } + new_brickinfo->caps = CAPS_BD; +#else + vg = NULL; /* Avoid compiler warnings when BD not enabled */ +#endif ret = gf_canonicalize_path (path); if (ret) goto out; @@ -745,6 +981,62 @@ out: return available; } +#ifdef HAVE_BD_XLATOR +/* + * Sets the tag of the format "trusted.glusterfs.volume-id:<uuid>" in + * the brick VG. It is used to avoid using same VG for another brick. + * @volume-id - gfid, @brick - brick info, @msg - Error message returned + * to the caller + */ +int +glusterd_bd_set_vg_tag (unsigned char *volume_id, glusterd_brickinfo_t *brick, + char *msg, int msg_size) +{ + lvm_t handle = NULL; + vg_t vg = NULL; + char *uuid = NULL; + int ret = -1; + + gf_asprintf (&uuid, "%s:%s", GF_XATTR_VOL_ID_KEY, + uuid_utoa (volume_id)); + if (!uuid) { + snprintf (msg, sizeof(*msg), "Could not allocate memory " + "for tag"); + return -1; + } + + handle = lvm_init (NULL); + if (!handle) { + snprintf (msg, sizeof(*msg), "lvm_init failed"); + goto out; + } + + vg = lvm_vg_open (handle, brick->vg, "w", 0); + if (!vg) { + snprintf (msg, sizeof(*msg), "Could not open VG %s", + brick->vg); + goto out; + } + + if (lvm_vg_add_tag (vg, uuid) < 0) { + snprintf (msg, sizeof(*msg), "Could not set tag %s for " + "VG %s", uuid, brick->vg); + goto out; + } + lvm_vg_write (vg); + ret = 0; +out: + GF_FREE (uuid); + + if (vg) + lvm_vg_close (vg); + if (handle) + lvm_quit (handle); + + return ret; +} +#endif + int glusterd_validate_and_create_brickpath (glusterd_brickinfo_t *brickinfo, uuid_t volume_id, char **op_errstr, @@ -827,9 +1119,17 @@ glusterd_validate_and_create_brickpath (glusterd_brickinfo_t *brickinfo, } } +#ifdef HAVE_BD_XLATOR + if (brickinfo->vg[0]) { + ret = glusterd_bd_set_vg_tag (volume_id, brickinfo, msg, + sizeof(msg)); + if (ret) + goto out; + } +#endif ret = glusterd_check_and_set_brick_xattr (brickinfo->hostname, brickinfo->path, volume_id, - op_errstr); + op_errstr, is_force); if (ret) goto out; @@ -987,6 +1287,42 @@ glusterd_volinfo_find_by_volume_id (uuid_t volume_id, glusterd_volinfo_t **volin return ret; } +int +glusterd_snap_volinfo_find_by_volume_id (uuid_t volume_id, + glusterd_volinfo_t **volinfo) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_volinfo_t *voliter = NULL; + glusterd_snap_t *snap = NULL; + glusterd_conf_t *priv = NULL; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (volinfo); + + if (uuid_is_null(volume_id)) { + gf_log (this->name, GF_LOG_WARNING, "Volume UUID is NULL"); + goto out; + } + + list_for_each_entry (snap, &priv->snapshots, snap_list) { + list_for_each_entry (voliter, &snap->volumes, vol_list) { + if (uuid_compare (volume_id, voliter->volume_id)) + continue; + *volinfo = voliter; + ret = 0; + goto out; + } + } + + gf_log (this->name, GF_LOG_WARNING, "Snap volume not found"); +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + int32_t glusterd_volinfo_find (char *volname, glusterd_volinfo_t **volinfo) { @@ -996,7 +1332,6 @@ glusterd_volinfo_find (char *volname, glusterd_volinfo_t **volinfo) glusterd_conf_t *priv = NULL; GF_ASSERT (volname); - this = THIS; GF_ASSERT (this); @@ -1005,7 +1340,8 @@ glusterd_volinfo_find (char *volname, glusterd_volinfo_t **volinfo) list_for_each_entry (tmp_volinfo, &priv->volumes, vol_list) { if (!strcmp (tmp_volinfo->volname, volname)) { - gf_log (this->name, GF_LOG_DEBUG, "Volume %s found", volname); + gf_log (this->name, GF_LOG_DEBUG, "Volume %s found", + volname); ret = 0; *volinfo = tmp_volinfo; break; @@ -1017,6 +1353,68 @@ glusterd_volinfo_find (char *volname, glusterd_volinfo_t **volinfo) } int32_t +glusterd_snap_volinfo_find (char *snap_volname, glusterd_snap_t *snap, + glusterd_volinfo_t **volinfo) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_conf_t *priv = NULL; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (snap); + GF_ASSERT (snap_volname); + + list_for_each_entry (snap_vol, &snap->volumes, vol_list) { + if (!strcmp (snap_vol->volname, snap_volname)) { + ret = 0; + *volinfo = snap_vol; + goto out; + } + } + + gf_log (this->name, GF_LOG_WARNING, "Snap volume %s not found", + snap_volname); +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_snap_volinfo_find_from_parent_volname (char *origin_volname, + glusterd_snap_t *snap, + glusterd_volinfo_t **volinfo) +{ + int32_t ret = -1; + xlator_t *this = NULL; + glusterd_volinfo_t *snap_vol = NULL; + glusterd_conf_t *priv = NULL; + + this = THIS; + priv = this->private; + GF_ASSERT (priv); + GF_ASSERT (snap); + GF_ASSERT (origin_volname); + + list_for_each_entry (snap_vol, &snap->volumes, vol_list) { + if (!strcmp (snap_vol->parent_volname, origin_volname)) { + ret = 0; + *volinfo = snap_vol; + goto out; + } + } + + gf_log (this->name, GF_LOG_DEBUG, "Snap volume not found(snap: %s, " + "origin-volume: %s", snap->snapname, origin_volname); + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t glusterd_service_stop (const char *service, char *pidfile, int sig, gf_boolean_t force_kill) { @@ -1096,10 +1494,9 @@ glusterd_set_brick_socket_filepath (glusterd_volinfo_t *volinfo, */ int32_t glusterd_brick_connect (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo) + glusterd_brickinfo_t *brickinfo, char *socketpath) { int ret = 0; - char socketpath[PATH_MAX] = {0}; char volume_id_str[64]; char *brickid = NULL; dict_t *options = NULL; @@ -1108,12 +1505,9 @@ glusterd_brick_connect (glusterd_volinfo_t *volinfo, GF_ASSERT (volinfo); GF_ASSERT (brickinfo); + GF_ASSERT (socketpath); if (brickinfo->rpc == NULL) { - glusterd_set_brick_socket_filepath (volinfo, brickinfo, - socketpath, - sizeof (socketpath)); - /* Setting frame-timeout to 10mins (600seconds). * Unix domain sockets ensures that the connection is reliable. * The default timeout of 30mins used for unreliable network @@ -1201,9 +1595,23 @@ glusterd_volume_start_glusterfs (glusterd_volinfo_t *volinfo, priv = this->private; GF_ASSERT (priv); + if (brickinfo->snap_status == -1) { + gf_log (this->name, GF_LOG_INFO, + "Snapshot is pending on %s:%s. " + "Hence not starting the brick", + brickinfo->hostname, + brickinfo->path); + ret = 0; + goto out; + } + ret = _mk_rundir_p (volinfo); if (ret) goto out; + + glusterd_set_brick_socket_filepath (volinfo, brickinfo, socketpath, + sizeof (socketpath)); + GLUSTERD_GET_BRICK_PIDFILE (pidfile, volinfo, brickinfo, priv); if (glusterd_is_service_running (pidfile, NULL)) goto connect; @@ -1241,8 +1649,15 @@ glusterd_volume_start_glusterfs (glusterd_volinfo_t *volinfo, runner_argprintf (&runner, "--log-file=%s", valgrind_logfile); } - snprintf (volfile, PATH_MAX, "%s.%s.%s", volinfo->volname, - brickinfo->hostname, exp_path); + if (volinfo->is_snap_volume) { + snprintf (volfile, PATH_MAX,"/%s/%s/%s.%s.%s", + GLUSTERD_VOL_SNAP_DIR_PREFIX, + volinfo->snapshot->snapname, volinfo->volname, + brickinfo->hostname, exp_path); + } else { + snprintf (volfile, PATH_MAX, "%s.%s.%s", volinfo->volname, + brickinfo->hostname, exp_path); + } if (volinfo->logdir) { snprintf (logfile, PATH_MAX, "%s/%s.log", @@ -1254,9 +1669,6 @@ glusterd_volume_start_glusterfs (glusterd_volinfo_t *volinfo, if (!brickinfo->logfile) brickinfo->logfile = gf_strdup (logfile); - glusterd_set_brick_socket_filepath (volinfo, brickinfo, socketpath, - sizeof (socketpath)); - (void) snprintf (glusterd_uuid, 1024, "*-posix.glusterd-uuid=%s", uuid_utoa (MY_UUID)); runner_add_args (&runner, SBIN_DIR"/glusterfsd", @@ -1304,9 +1716,13 @@ glusterd_volume_start_glusterfs (glusterd_volinfo_t *volinfo, brickinfo->rdma_port = rdma_port; connect: - ret = glusterd_brick_connect (volinfo, brickinfo); - if (ret) + ret = glusterd_brick_connect (volinfo, brickinfo, socketpath); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to connect to brick %s:%s on %s", + brickinfo->hostname, brickinfo->path, socketpath); goto out; + } out: return ret; } @@ -1372,10 +1788,10 @@ glusterd_volume_stop_glusterfs (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo, gf_boolean_t del_brick) { - xlator_t *this = NULL; - glusterd_conf_t *priv = NULL; - char pidfile[PATH_MAX] = {0,}; - int ret = 0; + xlator_t *this = NULL; + glusterd_conf_t *priv = NULL; + char pidfile[PATH_MAX] = {0,}; + int ret = 0; GF_ASSERT (volinfo); GF_ASSERT (brickinfo); @@ -1574,13 +1990,14 @@ glusterd_volume_compute_cksum (glusterd_volinfo_t *volinfo) snprintf (filepath, sizeof (filepath), "%s/%s", path, GLUSTERD_VOLUME_INFO_FILE); - snprintf (sort_filepath, sizeof (sort_filepath), "/tmp/%s.XXXXXX", - volinfo->volname); + snprintf (sort_filepath, sizeof (sort_filepath), + "/tmp/%s.XXXXXX", volinfo->volname); sort_fd = mkstemp (sort_filepath); if (sort_fd < 0) { gf_log (this->name, GF_LOG_ERROR, "Could not generate temp " - "file, reason: %s for volume: %s", strerror (errno), + "file, reason: %s for %s: %s", strerror (errno), + (volinfo->is_snap_volume)?"snap":"volume", volinfo->volname); goto out; } else { @@ -1621,7 +2038,6 @@ glusterd_volume_compute_cksum (glusterd_volinfo_t *volinfo) goto out; volinfo->cksum = cksum; - out: if (fd > 0) close (fd); @@ -1700,7 +2116,10 @@ glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo, glusterd_dict_ctx_t ctx = {0}; char *rebalance_id_str = NULL; char *rb_id_str = NULL; + xlator_t *this = NULL; + this = THIS; + GF_ASSERT (this); GF_ASSERT (dict); GF_ASSERT (volinfo); @@ -1715,6 +2134,15 @@ glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; + snprintf (key, sizeof (key), "volume%d.is_volume_restored", count); + ret = dict_set_int32 (dict, key, volinfo->is_volume_restored); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to set " + "is_volume_restored option for %s volume", + volinfo->volname); + goto out; + } + memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "volume%d.brick_count", count); ret = dict_set_int32 (dict, key, volinfo->brick_count); @@ -1769,6 +2197,20 @@ glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; + snprintf (key, sizeof (key), "volume%d.is_snap_volume", count); + ret = dict_set_uint32 (dict, key, volinfo->is_snap_volume); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Unable to set %s", key); + goto out; + } + + snprintf (key, sizeof (key), "volume%d.snap-max-hard-limit", count); + ret = dict_set_uint64 (dict, key, volinfo->snap_max_hard_limit); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Unable to set %s", key); + goto out; + } + volume_id_str = gf_strdup (uuid_utoa (volinfo->volume_id)); if (!volume_id_str) { ret = -1; @@ -1915,9 +2357,42 @@ glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo, if (ret) goto out; + snprintf (key, sizeof (key), "volume%d.brick%d.snap_status", + count, i); + ret = dict_set_int32 (dict, key, brickinfo->snap_status); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set snap_status for %s:%s", + brickinfo->hostname, + brickinfo->path); + goto out; + } + + snprintf (key, sizeof (key), "volume%d.brick%d.device_path", + count, i); + ret = dict_set_str (dict, key, brickinfo->device_path); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set snap_device for %s:%s", + brickinfo->hostname, + brickinfo->path); + goto out; + } + i++; } + /* Add volume op-versions to dict. This prevents volume inconsistencies + * in the cluster + */ + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.op-version", count); + ret = dict_set_int32 (dict, key, volinfo->op_version); + if (ret) + goto out; + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.client-op-version", count); + ret = dict_set_int32 (dict, key, volinfo->client_op_version); out: GF_FREE (volume_id_str); @@ -2304,7 +2779,7 @@ glusterd_do_volume_quorum_action (xlator_t *this, glusterd_volinfo_t *volinfo, gf_boolean_t meets_quorum) { glusterd_brickinfo_t *brickinfo = NULL; - glusterd_conf_t *conf = NULL; + glusterd_conf_t *conf = NULL; conf = this->private; if (volinfo->status != GLUSTERD_STATUS_STARTED) @@ -2422,6 +2897,8 @@ glusterd_import_new_brick (dict_t *vols, int32_t vol_count, { char key[512] = {0,}; int ret = -1; + int32_t snap_status = 0; + char *snap_device = NULL; char *hostname = NULL; char *path = NULL; glusterd_brickinfo_t *new_brickinfo = NULL; @@ -2449,12 +2926,31 @@ glusterd_import_new_brick (dict_t *vols, int32_t vol_count, goto out; } + snprintf (key, sizeof (key), "volume%d.brick%d.snap_status", + vol_count, brick_count); + ret = dict_get_int32 (vols, key, &snap_status); + if (ret) { + snprintf (msg, sizeof (msg), "%s missing in payload", key); + goto out; + } + + snprintf (key, sizeof (key), "volume%d.brick%d.device_path", + vol_count, brick_count); + ret = dict_get_str (vols, key, &snap_device); + if (ret) { + snprintf (msg, sizeof (msg), "%s missing in payload", key); + goto out; + } + ret = glusterd_brickinfo_new (&new_brickinfo); if (ret) goto out; strcpy (new_brickinfo->path, path); strcpy (new_brickinfo->hostname, hostname); + strcpy (new_brickinfo->device_path, snap_device); + new_brickinfo->snap_status = snap_status; + //peerinfo might not be added yet (void) glusterd_resolve_brick (new_brickinfo); ret = 0; @@ -2508,6 +3004,9 @@ glusterd_import_volinfo (dict_t *vols, int count, int rb_status = 0; char *rebalance_id_str = NULL; char *rb_id_str = NULL; + int op_version = 0; + int client_op_version = 0; + uint32_t is_snap_volume = 0; GF_ASSERT (vols); GF_ASSERT (volinfo); @@ -2519,6 +3018,22 @@ glusterd_import_volinfo (dict_t *vols, int count, goto out; } + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.is_snap_volume", count); + ret = dict_get_uint32 (vols, key, &is_snap_volume); + if (ret) { + snprintf (msg, sizeof (msg), "%s missing in payload for %s", + key, volname); + goto out; + } + + if (is_snap_volume == _gf_true) { + gf_log (THIS->name, GF_LOG_DEBUG, + "Not syncing snap volume %s", volname); + ret = 0; + goto out; + } + ret = glusterd_volinfo_new (&new_volinfo); if (ret) goto out; @@ -2644,6 +3159,25 @@ glusterd_import_volinfo (dict_t *vols, int count, goto out; } + new_volinfo->is_snap_volume = is_snap_volume; + + snprintf (key, sizeof (key), "volume%d.is_volume_restored", count); + ret = dict_get_uint32 (vols, key, &new_volinfo->is_volume_restored); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to get " + "is_volume_restored option for %s", + volname); + goto out; + } + + snprintf (key, sizeof (key), "volume%d.snap-max-hard-limit", count); + ret = dict_get_uint64 (vols, key, &new_volinfo->snap_max_hard_limit); + if (ret) { + snprintf (msg, sizeof (msg), "%s missing in payload for %s", + key, volname); + goto out; + } + memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "volume%d.rebalance", count); ret = dict_get_uint32 (vols, key, &new_volinfo->rebal.defrag_cmd); @@ -2734,6 +3268,40 @@ glusterd_import_volinfo (dict_t *vols, int count, ret = glusterd_import_friend_volume_opts (vols, count, new_volinfo); if (ret) goto out; + + /* Import the volume's op-versions if available else set it to 1. + * Not having op-versions implies this informtation was obtained from a + * op-version 1 friend (gluster-3.3), ergo the cluster is at op-version + * 1 and all volumes are at op-versions 1. + * + * Either both the volume op-versions should be absent or both should be + * present. Only one being present is a failure + */ + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.op-version", count); + ret = dict_get_int32 (vols, key, &op_version); + if (ret) + ret = 0; + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "volume%d.client-op-version", count); + ret = dict_get_int32 (vols, key, &client_op_version); + if (ret) + ret = 0; + + if (op_version && client_op_version) { + new_volinfo->op_version = op_version; + new_volinfo->client_op_version = client_op_version; + } else if (((op_version == 0) && (client_op_version != 0)) || + ((op_version != 0) && (client_op_version == 0))) { + ret = -1; + gf_log ("glusterd", GF_LOG_ERROR, + "Only one volume op-version found"); + goto out; + } else { + new_volinfo->op_version = 1; + new_volinfo->client_op_version = 1; + } + ret = glusterd_import_bricks (vols, count, new_volinfo); if (ret) goto out; @@ -2892,6 +3460,12 @@ glusterd_import_friend_volume (dict_t *vols, size_t count) if (ret) goto out; + if (!new_volinfo) { + gf_log (this->name, GF_LOG_DEBUG, + "Not importing snap volume"); + goto out; + } + ret = glusterd_volinfo_find (new_volinfo->volname, &old_volinfo); if (0 == ret) { (void) glusterd_delete_stale_volume (old_volinfo, new_volinfo); @@ -2901,8 +3475,6 @@ glusterd_import_friend_volume (dict_t *vols, size_t count) (void) glusterd_start_bricks (new_volinfo); } - gd_update_volume_op_versions (new_volinfo); - ret = glusterd_store_volinfo (new_volinfo, GLUSTERD_VOLINFO_VER_AC_NONE); ret = glusterd_create_volfiles_and_notify_services (new_volinfo); if (ret) @@ -3710,6 +4282,11 @@ glusterd_reconfigure_nfs () int ret = -1; gf_boolean_t identical = _gf_false; + /* + * Check both OLD and NEW volfiles, if they are SAME by size + * and cksum i.e. "character-by-character". If YES, then + * NOTHING has been changed, just return. + */ ret = glusterd_check_nfs_volfile_identical (&identical); if (ret) goto out; @@ -3719,6 +4296,31 @@ glusterd_reconfigure_nfs () goto out; } + /* + * They are not identical. Find out if the topology is changed + * OR just the volume options. If just the options which got + * changed, then inform the xlator to reconfigure the options. + */ + identical = _gf_false; /* RESET the FLAG */ + ret = glusterd_check_nfs_topology_identical (&identical); + if (ret) + goto out; + + /* Topology is not changed, but just the options. But write the + * options to NFS volfile, so that NFS will be reconfigured. + */ + if (identical) { + ret = glusterd_create_nfs_volfile(); + if (ret == 0) {/* Only if above PASSES */ + ret = glusterd_fetchspec_notify (THIS); + } + goto out; + } + + /* + * NFS volfile's topology has been changed. NFS server needs + * to be RESTARTED to ACT on the changed volfile. + */ ret = glusterd_check_generate_start_nfs (); out: @@ -3950,20 +4552,42 @@ out: int glusterd_restart_bricks (glusterd_conf_t *conf) { + int ret = 0; glusterd_volinfo_t *volinfo = NULL; glusterd_brickinfo_t *brickinfo = NULL; + glusterd_snap_t *snap = NULL; gf_boolean_t start_nodesvcs = _gf_false; - int ret = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); list_for_each_entry (volinfo, &conf->volumes, vol_list) { if (volinfo->status != GLUSTERD_STATUS_STARTED) continue; start_nodesvcs = _gf_true; + gf_log (this->name, GF_LOG_DEBUG, "starting the volume %s", + volinfo->volname); list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { glusterd_brick_start (volinfo, brickinfo, _gf_false); } } + list_for_each_entry (snap, &conf->snapshots, snap_list) { + list_for_each_entry (volinfo, &snap->volumes, vol_list) { + if (volinfo->status != GLUSTERD_STATUS_STARTED) + continue; + start_nodesvcs = _gf_true; + gf_log (this->name, GF_LOG_DEBUG, "starting the snap " + "volume %s", volinfo->volname); + list_for_each_entry (brickinfo, &volinfo->bricks, + brick_list) { + glusterd_brick_start (volinfo, brickinfo, + _gf_false); + } + } + } + if (start_nodesvcs) glusterd_nodesvcs_handle_graph_change (NULL); @@ -4129,7 +4753,7 @@ out: } #ifdef GF_LINUX_HOST_OS -static int +int glusterd_get_brick_root (char *path, char **mount_point) { char *ptr = NULL; @@ -4297,6 +4921,31 @@ glusterd_add_inode_size_to_dict (dict_t *dict, int count) return ret; } +struct mntent * +glusterd_get_mnt_entry_info (char *mnt_pt, FILE *mtab) +{ + struct mntent *entry = NULL; + + mtab = setmntent (_PATH_MOUNTED, "r"); + if (!mtab) + goto out; + + entry = getmntent (mtab); + + while (1) { + if (!entry) + goto out; + + if (!strcmp (entry->mnt_dir, mnt_pt) && + strcmp (entry->mnt_type, "rootfs")) + break; + entry = getmntent (mtab); + } + +out: + return entry; +} + static int glusterd_add_brick_mount_details (glusterd_brickinfo_t *brickinfo, dict_t *dict, int count) @@ -4308,8 +4957,8 @@ glusterd_add_brick_mount_details (glusterd_brickinfo_t *brickinfo, char *fs_name = NULL; char *mnt_options = NULL; char *device = NULL; - FILE *mtab = NULL; struct mntent *entry = NULL; + FILE *mtab = NULL; snprintf (base_key, sizeof (base_key), "brick%d", count); @@ -4317,25 +4966,12 @@ glusterd_add_brick_mount_details (glusterd_brickinfo_t *brickinfo, if (ret) goto out; - mtab = setmntent (_PATH_MOUNTED, "r"); - if (!mtab) { + entry = glusterd_get_mnt_entry_info (mnt_pt, mtab); + if (!entry) { ret = -1; goto out; } - entry = getmntent (mtab); - - while (1) { - if (!entry) { - ret = -1; - goto out; - } - if (!strcmp (entry->mnt_dir, mnt_pt) && - strcmp (entry->mnt_type, "rootfs")) - break; - entry = getmntent (mtab); - } - /* get device file */ memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "%s.device", base_key); @@ -4368,6 +5004,45 @@ glusterd_add_brick_mount_details (glusterd_brickinfo_t *brickinfo, return ret; } + +char* +glusterd_get_brick_mount_details (glusterd_brickinfo_t *brickinfo) +{ + int ret = -1; + char *mnt_pt = NULL; + char *device = NULL; + FILE *mtab = NULL; + struct mntent *entry = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (brickinfo); + + ret = glusterd_get_brick_root (brickinfo->path, &mnt_pt); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get mount point " + "for %s brick", brickinfo->path); + goto out; + } + + entry = glusterd_get_mnt_entry_info (mnt_pt, mtab); + if (NULL == entry) { + gf_log (this->name, GF_LOG_ERROR, "Failed to get mnt entry " + "for %s mount path", mnt_pt); + goto out; + } + + /* get the fs_name/device */ + device = gf_strdup (entry->mnt_fsname); + +out: + if (NULL != mtab) { + endmntent (mtab); + } + + return device; +} #endif int @@ -4728,6 +5403,7 @@ glusterd_hostname_to_uuid (char *hostname, uuid_t uuid) uuid_copy (uuid, MY_UUID); ret = 0; } else { + ret = 0; goto out; } } else { @@ -5062,11 +5738,12 @@ out: int glusterd_check_and_set_brick_xattr (char *host, char *path, uuid_t uuid, - char **op_errstr) + char **op_errstr, gf_boolean_t is_force) { int ret = -1; char msg[2048] = {0,}; gf_boolean_t in_use = _gf_false; + int flags = 0; /* Check for xattr support in backend fs */ ret = sys_lsetxattr (path, "trusted.glusterfs.test", @@ -5086,13 +5763,17 @@ glusterd_check_and_set_brick_xattr (char *host, char *path, uuid_t uuid, if (ret) goto out; - if (in_use) { + if (in_use && !is_force) { ret = -1; goto out; } + + if (!is_force) + flags = XATTR_CREATE; + ret = sys_lsetxattr (path, GF_XATTR_VOL_ID_KEY, uuid, 16, - XATTR_CREATE); + flags); if (ret) { snprintf (msg, sizeof (msg), "Failed to set extended " "attributes %s, reason: %s", @@ -5604,17 +6285,38 @@ out: } int32_t -glusterd_recreate_bricks (glusterd_conf_t *conf) +glusterd_recreate_volfiles (glusterd_conf_t *conf) { glusterd_volinfo_t *volinfo = NULL; int ret = 0; + int op_ret = 0; GF_ASSERT (conf); list_for_each_entry (volinfo, &conf->volumes, vol_list) { ret = generate_brick_volfiles (volinfo); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, "Failed to " + "regenerate brick volfiles for %s", + volinfo->volname); + op_ret = ret; + } + ret = generate_client_volfiles (volinfo, GF_CLIENT_TRUSTED); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, "Failed to " + "regenerate trusted client volfiles for %s", + volinfo->volname); + op_ret = ret; + } + ret = generate_client_volfiles (volinfo, GF_CLIENT_OTHER); + if (ret) { + gf_log ("glusterd", GF_LOG_ERROR, "Failed to " + "regenerate client volfiles for %s", + volinfo->volname); + op_ret = ret; + } } - return ret; + return op_ret; } int32_t @@ -5624,7 +6326,7 @@ glusterd_handle_upgrade_downgrade (dict_t *options, glusterd_conf_t *conf) char *type = NULL; gf_boolean_t upgrade = _gf_false; gf_boolean_t downgrade = _gf_false; - gf_boolean_t regenerate_brick_volfiles = _gf_false; + gf_boolean_t regenerate_volfiles = _gf_false; gf_boolean_t terminate = _gf_false; ret = dict_get_str (options, "upgrade", &type); @@ -5637,7 +6339,7 @@ glusterd_handle_upgrade_downgrade (dict_t *options, glusterd_conf_t *conf) goto out; } if (_gf_true == upgrade) - regenerate_brick_volfiles = _gf_true; + regenerate_volfiles = _gf_true; } ret = dict_get_str (options, "downgrade", &type); @@ -5662,8 +6364,8 @@ glusterd_handle_upgrade_downgrade (dict_t *options, glusterd_conf_t *conf) ret = 0; else terminate = _gf_true; - if (regenerate_brick_volfiles) { - ret = glusterd_recreate_bricks (conf); + if (regenerate_volfiles) { + ret = glusterd_recreate_volfiles (conf); } out: if (terminate && (ret == 0)) @@ -6215,6 +6917,68 @@ glusterd_defrag_volume_status_update (glusterd_volinfo_t *volinfo, } int +glusterd_check_topology_identical (const char *filename1, + const char *filename2, + gf_boolean_t *identical) +{ + int ret = -1; /* FAILURE */ + xlator_t *this = NULL; + FILE *fp1 = NULL; + FILE *fp2 = NULL; + glusterfs_graph_t *grph1 = NULL; + glusterfs_graph_t *grph2 = NULL; + + if ((!filename1) || (!filename2) || (!identical)) + goto out; + + this = THIS; + + errno = 0; /* RESET the errno */ + + /* fopen() the volfile1 to create the graph */ + fp1 = fopen (filename1, "r"); + if (fp1 == NULL) { + gf_log (this->name, GF_LOG_ERROR, "fopen() on file: %s failed " + "(%s)", filename1, strerror (errno)); + goto out; + } + + /* fopen() the volfile2 to create the graph */ + fp2 = fopen (filename2, "r"); + if (fp2 == NULL) { + gf_log (this->name, GF_LOG_ERROR, "fopen() on file: %s failed " + "(%s)", filename2, strerror (errno)); + goto out; + } + + /* create the graph for filename1 */ + grph1 = glusterfs_graph_construct(fp1); + if (grph1 == NULL) + goto out; + + /* create the graph for filename2 */ + grph2 = glusterfs_graph_construct(fp2); + if (grph2 == NULL) + goto out; + + /* compare the graph topology */ + *identical = is_graph_topology_equal(grph1, grph2); + ret = 0; /* SUCCESS */ +out: + if (fp1) + fclose(fp1); + if (fp2) + fclose(fp2); + if (grph1) + glusterfs_graph_destroy(grph1); + if (grph2) + glusterfs_graph_destroy(grph2); + + gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); + return ret; +} + +int glusterd_check_files_identical (char *filename1, char *filename2, gf_boolean_t *identical) { @@ -6474,6 +7238,7 @@ glusterd_gsync_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict, char *op_errstr) { dict_t *ctx = NULL; int ret = 0; + char *conf_path = NULL; if (aggr) { ctx = aggr; @@ -6495,6 +7260,17 @@ glusterd_gsync_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict, char *op_errstr) ret = glusterd_append_gsync_status (ctx, rsp_dict); if (ret) goto out; + + ret = dict_get_str (rsp_dict, "conf_path", &conf_path); + if (!ret && conf_path) { + ret = dict_set_dynstr (ctx, "conf_path", + gf_strdup(conf_path)); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to store conf path."); + goto out; + } + } } if ((op_errstr) && (strcmp ("", op_errstr))) { ret = dict_set_dynstr (ctx, "errstr", gf_strdup(op_errstr)); @@ -6715,7 +7491,7 @@ glusterd_volume_status_copy_to_op_ctx_dict (dict_t *aggr, dict_t *rsp_dict) if (ret) goto out; - if (cmd & GF_CLI_STATUS_ALL && is_origin_glusterd ()) { + if (cmd & GF_CLI_STATUS_ALL && is_origin_glusterd (ctx_dict)) { ret = dict_get_int32 (rsp_dict, "vol_count", &vol_count); if (ret == 0) { ret = dict_set_int32 (ctx_dict, "vol_count", @@ -6737,6 +7513,12 @@ glusterd_volume_status_copy_to_op_ctx_dict (dict_t *aggr, dict_t *rsp_dict) } } + if ((cmd & GF_CLI_STATUS_TASKS) != 0) { + dict_copy (rsp_dict, aggr); + ret = 0; + goto out; + } + ret = dict_get_int32 (rsp_dict, "count", &rsp_node_count); if (ret) { ret = 0; //no bricks in the rsp @@ -6970,6 +7752,279 @@ out: } int +glusterd_snap_config_use_rsp_dict (dict_t *dst, dict_t *src) +{ + char buf[PATH_MAX] = ""; + char *volname = NULL; + int ret = -1; + int config_command = 0; + uint64_t i = 0; + uint64_t value = 0; + uint64_t voldisplaycount = 0; + + if (!dst || !src) { + gf_log ("", GF_LOG_ERROR, "Source or Destination " + "dict is empty."); + goto out; + } + + ret = dict_get_int32 (dst, "config-command", &config_command); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "failed to get config-command type"); + goto out; + } + + switch (config_command) { + case GF_SNAP_CONFIG_DISPLAY: + ret = dict_get_uint64 (src, "snap-max-hard-limit", &value); + if (!ret) { + ret = dict_set_uint64 (dst, "snap-max-hard-limit", value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set snap_max_hard_limit"); + goto out; + } + } else { + /* Received dummy response from other nodes */ + ret = 0; + goto out; + } + + ret = dict_get_uint64 (src, "snap-max-soft-limit", &value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get snap_max_soft_limit"); + goto out; + } + + ret = dict_set_uint64 (dst, "snap-max-soft-limit", value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set snap_max_soft_limit"); + goto out; + } + + ret = dict_get_uint64 (src, "voldisplaycount", + &voldisplaycount); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get voldisplaycount"); + goto out; + } + + ret = dict_set_uint64 (dst, "voldisplaycount", + voldisplaycount); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set voldisplaycount"); + goto out; + } + + for (i = 0; i < voldisplaycount; i++) { + snprintf (buf, sizeof(buf), "volume%ld-volname", i); + ret = dict_get_str (src, buf, &volname); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get %s", buf); + goto out; + } + ret = dict_set_str (dst, buf, volname); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-hard-limit", i); + ret = dict_get_uint64 (src, buf, &value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get %s", buf); + goto out; + } + ret = dict_set_uint64 (dst, buf, value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-active-hard-limit", i); + ret = dict_get_uint64 (src, buf, &value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get %s", buf); + goto out; + } + ret = dict_set_uint64 (dst, buf, value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set %s", buf); + goto out; + } + + snprintf (buf, sizeof(buf), + "volume%ld-snap-max-soft-limit", i); + ret = dict_get_uint64 (src, buf, &value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get %s", buf); + goto out; + } + ret = dict_set_uint64 (dst, buf, value); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set %s", buf); + goto out; + } + } + + break; + default: + break; + } + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +/* Aggregate missed_snap_counts from different nodes and save it * + * in the req_dict of the originator node */ +int +glusterd_snap_create_use_rsp_dict (dict_t *dst, dict_t *src) +{ + char *buf = NULL; + char *tmp_str = NULL; + char name_buf[PATH_MAX] = ""; + int32_t i = -1; + int32_t ret = -1; + int32_t src_missed_snap_count = -1; + int32_t dst_missed_snap_count = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + if (!dst || !src) { + gf_log (this->name, GF_LOG_ERROR, "Source or Destination " + "dict is empty."); + goto out; + } + + ret = dict_get_int32 (src, "missed_snap_count", + &src_missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, "No missed snaps"); + ret = 0; + goto out; + } + + ret = dict_get_int32 (dst, "missed_snap_count", + &dst_missed_snap_count); + if (ret) { + /* Initialize dst_missed_count for the first time */ + dst_missed_snap_count = 0; + } + + for (i = 0; i < src_missed_snap_count; i++) { + snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d", + i); + ret = dict_get_str (src, name_buf, &buf); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to fetch %s", name_buf); + goto out; + } + + snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d", + dst_missed_snap_count); + + tmp_str = gf_strdup (buf); + if (!tmp_str) { + ret = -1; + goto out; + } + + ret = dict_set_dynstr (dst, name_buf, tmp_str); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set %s", name_buf); + goto out; + } + + tmp_str = NULL; + dst_missed_snap_count++; + } + + ret = dict_set_int32 (dst, "missed_snap_count", dst_missed_snap_count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to set dst_missed_snap_count"); + goto out; + } + +out: + if (ret && tmp_str) + GF_FREE(tmp_str); + + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int +glusterd_snap_use_rsp_dict (dict_t *dst, dict_t *src) +{ + int ret = -1; + int32_t snap_command = 0; + + if (!dst || !src) { + gf_log ("", GF_LOG_ERROR, "Source or Destination " + "dict is empty."); + goto out; + } + + ret = dict_get_int32 (dst, "type", &snap_command); + if (ret) { + gf_log ("", GF_LOG_ERROR, "unable to get the type of " + "the snapshot command"); + goto out; + } + + switch (snap_command) { + case GF_SNAP_OPTION_TYPE_CREATE: + case GF_SNAP_OPTION_TYPE_DELETE: + ret = glusterd_snap_create_use_rsp_dict (dst, src); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to use rsp dict"); + goto out; + } + break; + case GF_SNAP_OPTION_TYPE_CONFIG: + ret = glusterd_snap_config_use_rsp_dict (dst, src); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to use rsp dict"); + goto out; + } + break; + default: + // copy the response dictinary's contents to the dict to be + // sent back to the cli + dict_copy (src, dst); + break; + } + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int glusterd_sys_exec_output_rsp_dict (dict_t *dst, dict_t *src) { char output_name[PATH_MAX] = ""; @@ -7197,6 +8252,77 @@ out: } int +_heal_volume_add_shd_rsp_of_statistics (dict_t *this, char *key, data_t + *value, void *data) +{ + char new_key[256] = {0,}; + char int_str[16] = {0,}; + char key_begin_string[128] = {0,}; + data_t *new_value = NULL; + char *rxl_end = NULL; + char *rxl_child_end = NULL; + glusterd_volinfo_t *volinfo = NULL; + char *key_begin_str = NULL; + int rxl_id = 0; + int rxl_child_id = 0; + int brick_id = 0; + int int_len = 0; + int ret = 0; + glusterd_heal_rsp_conv_t *rsp_ctx = NULL; + glusterd_brickinfo_t *brickinfo = NULL; + + rsp_ctx = data; + key_begin_str = strchr (key, '-'); + if (!key_begin_str) + goto out; + + int_len = strlen (key) - strlen (key_begin_str); + strncpy (key_begin_string, key, int_len); + key_begin_string[int_len] = '\0'; + + rxl_end = strchr (key_begin_str + 1, '-'); + if (!rxl_end) + goto out; + + int_len = strlen (key_begin_str) - strlen (rxl_end) - 1; + strncpy (int_str, key_begin_str + 1, int_len); + int_str[int_len] = '\0'; + ret = gf_string2int (int_str, &rxl_id); + if (ret) + goto out; + + + rxl_child_end = strchr (rxl_end + 1, '-'); + if (!rxl_child_end) + goto out; + + int_len = strlen (rxl_end) - strlen (rxl_child_end) - 1; + strncpy (int_str, rxl_end + 1, int_len); + int_str[int_len] = '\0'; + ret = gf_string2int (int_str, &rxl_child_id); + if (ret) + goto out; + + volinfo = rsp_ctx->volinfo; + brick_id = rxl_id * volinfo->replica_count + rxl_child_id; + + brickinfo = glusterd_get_brickinfo_by_position (volinfo, brick_id); + if (!brickinfo) + goto out; + if (!glusterd_is_local_brick (rsp_ctx->this, volinfo, brickinfo)) + goto out; + + new_value = data_copy (value); + snprintf (new_key, sizeof (new_key), "%s-%d%s", key_begin_string, + brick_id, rxl_child_end); + dict_set (rsp_ctx->dict, new_key, new_value); + +out: + return 0; + +} + +int glusterd_heal_volume_brick_rsp (dict_t *req_dict, dict_t *rsp_dict, dict_t *op_ctx, char **op_errstr) { @@ -7204,6 +8330,7 @@ glusterd_heal_volume_brick_rsp (dict_t *req_dict, dict_t *rsp_dict, glusterd_heal_rsp_conv_t rsp_ctx = {0}; char *volname = NULL; glusterd_volinfo_t *volinfo = NULL; + int heal_op = -1; GF_ASSERT (rsp_dict); GF_ASSERT (op_ctx); @@ -7215,6 +8342,13 @@ glusterd_heal_volume_brick_rsp (dict_t *req_dict, dict_t *rsp_dict, goto out; } + ret = dict_get_int32 (req_dict, "heal-op", &heal_op); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to get heal_op"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); if (ret) @@ -7223,7 +8357,12 @@ glusterd_heal_volume_brick_rsp (dict_t *req_dict, dict_t *rsp_dict, rsp_ctx.dict = op_ctx; rsp_ctx.volinfo = volinfo; rsp_ctx.this = THIS; - dict_foreach (rsp_dict, _heal_volume_add_shd_rsp, &rsp_ctx); + if (heal_op == GF_AFR_OP_STATISTICS) + dict_foreach (rsp_dict, _heal_volume_add_shd_rsp_of_statistics, + &rsp_ctx); + else + dict_foreach (rsp_dict, _heal_volume_add_shd_rsp, &rsp_ctx); + out: return ret; @@ -7426,16 +8565,31 @@ glusterd_handle_node_rsp (dict_t *req_dict, void *pending_entry, * time a lock_owner is set */ gf_boolean_t -is_origin_glusterd () +is_origin_glusterd (dict_t *dict) { - int ret = 0; - uuid_t lock_owner = {0,}; + gf_boolean_t ret = _gf_false; + uuid_t lock_owner = {0,}; + uuid_t *originator_uuid = NULL; - ret = glusterd_get_lock_owner (&lock_owner); - if (ret) - return _gf_false; + GF_ASSERT (dict); + + ret = dict_get_bin (dict, "originator_uuid", + (void **) &originator_uuid); + if (ret) { + /* If not originator_uuid has been set, then the command + * has been originated from a glusterd running on older version + * Hence fetching the lock owner */ + ret = glusterd_get_lock_owner (&lock_owner); + if (ret) { + ret = _gf_false; + goto out; + } + ret = !uuid_compare (MY_UUID, lock_owner); + } else + ret = !uuid_compare (MY_UUID, *originator_uuid); - return (uuid_compare (MY_UUID, lock_owner) == 0); +out: + return ret; } int @@ -7619,3 +8773,122 @@ out: gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } + +gf_boolean_t +glusterd_is_status_tasks_op (glusterd_op_t op, dict_t *dict) +{ + int ret = -1; + uint32_t cmd = GF_CLI_STATUS_NONE; + gf_boolean_t is_status_tasks = _gf_false; + + if (op != GD_OP_STATUS_VOLUME) + goto out; + + ret = dict_get_uint32 (dict, "cmd", &cmd); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Failed to get opcode"); + goto out; + } + + if (cmd & GF_CLI_STATUS_TASKS) + is_status_tasks = _gf_true; + +out: + return is_status_tasks; +} + +int +glusterd_compare_snap_time(struct list_head *list1, struct list_head *list2) +{ + glusterd_snap_t *snap1 = NULL; + glusterd_snap_t *snap2 = NULL; + double diff_time = 0; + + GF_ASSERT (list1); + GF_ASSERT (list2); + + snap1 = list_entry(list1, glusterd_snap_t, snap_list); + snap2 = list_entry(list2, glusterd_snap_t, snap_list); + diff_time = difftime(snap1->time_stamp, snap2->time_stamp); + + return ((int)diff_time); +} + +int +glusterd_compare_snap_vol_time(struct list_head *list1, struct list_head *list2) +{ + glusterd_volinfo_t *snapvol1 = NULL; + glusterd_volinfo_t *snapvol2 = NULL; + double diff_time = 0; + + GF_ASSERT (list1); + GF_ASSERT (list2); + + snapvol1 = list_entry(list1, glusterd_volinfo_t, snapvol_list); + snapvol2 = list_entry(list2, glusterd_volinfo_t, snapvol_list); + diff_time = difftime(snapvol1->snapshot->time_stamp, + snapvol2->snapshot->time_stamp); + + return ((int)diff_time); +} + +int32_t +glusterd_missed_snapinfo_new (glusterd_missed_snap_info **missed_snapinfo) +{ + glusterd_missed_snap_info *new_missed_snapinfo = NULL; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (missed_snapinfo); + + new_missed_snapinfo = GF_CALLOC (1, sizeof(*new_missed_snapinfo), + gf_gld_mt_missed_snapinfo_t); + + if (!new_missed_snapinfo) + goto out; + + new_missed_snapinfo->node_snap_info = NULL; + INIT_LIST_HEAD (&new_missed_snapinfo->missed_snaps); + INIT_LIST_HEAD (&new_missed_snapinfo->snap_ops); + + *missed_snapinfo = new_missed_snapinfo; + + ret = 0; + +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_missed_snap_op_new (glusterd_snap_op_t **snap_op) +{ + glusterd_snap_op_t *new_snap_op = NULL; + int32_t ret = -1; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (snap_op); + + new_snap_op = GF_CALLOC (1, sizeof(*new_snap_op), + gf_gld_mt_missed_snapinfo_t); + + if (!new_snap_op) + goto out; + + new_snap_op->brick_path = NULL; + new_snap_op->brick_num = -1; + new_snap_op->op = -1; + new_snap_op->status = -1; + INIT_LIST_HEAD (&new_snap_op->snap_ops_list); + + *snap_op = new_snap_op; + + ret = 0; +out: + gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret); + return ret; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h index 4d1dde7c8..56bb799bf 100644 --- a/xlators/mgmt/glusterd/src/glusterd-utils.h +++ b/xlators/mgmt/glusterd/src/glusterd-utils.h @@ -81,6 +81,11 @@ glusterd_submit_request (struct rpc_clnt *rpc, void *req, int32_t glusterd_volinfo_new (glusterd_volinfo_t **volinfo); +int32_t +glusterd_volinfo_dup (glusterd_volinfo_t *volinfo, + glusterd_volinfo_t **dup_volinfo, + gf_boolean_t set_userauth); + char * glusterd_auth_get_username (glusterd_volinfo_t *volinfo); @@ -115,11 +120,23 @@ int32_t glusterd_peer_hostname_new (char *hostname, glusterd_peer_hostname_t **name); int32_t +glusterd_snap_volinfo_find (char *volname, glusterd_snap_t *snap, + glusterd_volinfo_t **volinfo); +int32_t +glusterd_snap_volinfo_find_from_parent_volname (char *origin_volname, + glusterd_snap_t *snap, + glusterd_volinfo_t **volinfo); + +int32_t glusterd_volinfo_find (char *volname, glusterd_volinfo_t **volinfo); int glusterd_volinfo_find_by_volume_id (uuid_t volume_id, glusterd_volinfo_t **volinfo); +int +glusterd_snap_volinfo_find_by_volume_id (uuid_t volume_id, + glusterd_volinfo_t **volinfo); + int32_t glusterd_service_stop(const char *service, char *pidfile, int sig, gf_boolean_t force_kill); @@ -282,6 +299,7 @@ glusterd_is_defrag_on (glusterd_volinfo_t *volinfo); int32_t glusterd_volinfo_bricks_delete (glusterd_volinfo_t *volinfo); + int glusterd_friend_find_by_uuid (uuid_t uuid, glusterd_peerinfo_t **peerinfo); @@ -318,7 +336,7 @@ glusterd_rb_check_bricks (glusterd_volinfo_t *volinfo, int glusterd_check_and_set_brick_xattr (char *host, char *path, uuid_t uuid, - char **op_errstr); + char **op_errstr, gf_boolean_t is_force); int glusterd_validate_and_create_brickpath (glusterd_brickinfo_t *brickinfo, @@ -351,7 +369,7 @@ gf_boolean_t glusterd_peerinfo_is_uuid_unknown (glusterd_peerinfo_t *peerinfo); int32_t glusterd_brick_connect (glusterd_volinfo_t *volinfo, - glusterd_brickinfo_t *brickinfo); + glusterd_brickinfo_t *brickinfo, char *socketpath); int32_t glusterd_brick_disconnect (glusterd_brickinfo_t *brickinfo); int32_t @@ -359,12 +377,16 @@ glusterd_delete_volume (glusterd_volinfo_t *volinfo); int32_t glusterd_delete_brick (glusterd_volinfo_t* volinfo, glusterd_brickinfo_t *brickinfo); + int32_t glusterd_delete_all_bricks (glusterd_volinfo_t* volinfo); + int glusterd_spawn_daemons (void *opaque); + int glusterd_restart_gsyncds (glusterd_conf_t *conf); + int glusterd_start_gsync (glusterd_volinfo_t *master_vol, char *slave, char *path_list, char *conf_path, @@ -456,6 +478,12 @@ glusterd_defrag_volume_status_update (glusterd_volinfo_t *volinfo, int glusterd_check_files_identical (char *filename1, char *filename2, gf_boolean_t *identical); + +int +glusterd_check_topology_identical (const char *filename1, + const char *filename2, + gf_boolean_t *identical); + void glusterd_volinfo_reset_defrag_stats (glusterd_volinfo_t *volinfo); int @@ -479,6 +507,8 @@ int glusterd_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict); int glusterd_sys_exec_output_rsp_dict (dict_t *aggr, dict_t *rsp_dict); +int +glusterd_snap_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict); int32_t glusterd_handle_node_rsp (dict_t *req_ctx, void *pending_entry, glusterd_op_t op, dict_t *rsp_dict, dict_t *op_ctx, @@ -504,7 +534,7 @@ glusterd_are_vol_all_peers_up (glusterd_volinfo_t *volinfo, * time a lock_owner is set */ gf_boolean_t -is_origin_glusterd (); +is_origin_glusterd (dict_t *dict); gf_boolean_t glusterd_is_quorum_changed (dict_t *options, char *option, char *value); @@ -553,7 +583,7 @@ glusterd_are_vol_all_peers_up (glusterd_volinfo_t *volinfo, int glusterd_get_slave_details_confpath (glusterd_volinfo_t *volinfo, dict_t *dict, char **slave_ip, char **slave_vol, - char **conf_path); + char **conf_path, char **op_errstr); int glusterd_check_restart_gsync_session (glusterd_volinfo_t *volinfo, char *slave, @@ -565,4 +595,40 @@ glusterd_check_gsync_running_local (char *master, char *slave, char *conf_path, gf_boolean_t *is_run); +gf_boolean_t +glusterd_is_status_tasks_op (glusterd_op_t op, dict_t *dict); + +#ifdef GF_LINUX_HOST_OS +char* +glusterd_get_brick_mount_details (glusterd_brickinfo_t *brickinfo); +struct mntent * +glusterd_get_mnt_entry_info (char *mnt_pt, FILE *mtab); +int +glusterd_get_brick_root (char *path, char **mount_point); +#endif //LINUX_HOST + +int +glusterd_compare_snap_time(struct list_head *, struct list_head *); + +int +glusterd_compare_snap_vol_time(struct list_head *, struct list_head *); + +int32_t +glusterd_snap_volinfo_restore (dict_t *rsp_dict, + glusterd_volinfo_t *new_volinfo, + glusterd_volinfo_t *snap_volinfo); +int32_t +glusterd_lvm_snapshot_remove (dict_t *rsp_dict, glusterd_volinfo_t *snap_vol); + +int32_t +glusterd_missed_snapinfo_new (glusterd_missed_snap_info **missed_snapinfo); + +int32_t +glusterd_missed_snap_op_new (glusterd_snap_op_t **snap_op); + +int32_t +glusterd_add_missed_snaps_to_dict (dict_t *rsp_dict, char *snap_uuid, + glusterd_brickinfo_t *brickinfo, + int32_t brick_number, int32_t op); + #endif diff --git a/xlators/mgmt/glusterd/src/glusterd-volgen.c b/xlators/mgmt/glusterd/src/glusterd-volgen.c index 06a2d37d5..6f3c69e7d 100644 --- a/xlators/mgmt/glusterd/src/glusterd-volgen.c +++ b/xlators/mgmt/glusterd/src/glusterd-volgen.c @@ -28,6 +28,8 @@ #include "logging.h" #include "dict.h" #include "graph-utils.h" +#include "glusterd-store.h" +#include "glusterd-hooks.h" #include "trie.h" #include "glusterd-mem-types.h" #include "cli1-xdr.h" @@ -594,6 +596,8 @@ get_server_xlator (char *xlator) subvol = GF_XLATOR_MARKER; if (strcmp (xlator, "io-stats") == 0) subvol = GF_XLATOR_IO_STATS; + if (strcmp (xlator, "bd") == 0) + subvol = GF_XLATOR_BD; return subvol; } @@ -1419,8 +1423,6 @@ server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, char *password = NULL; char index_basepath[PATH_MAX] = {0}; char key[1024] = {0}; - char *vgname = NULL; - char *vg = NULL; glusterd_brickinfo_t *brickinfo = NULL; char changelog_basepath[PATH_MAX] = {0,}; @@ -1441,47 +1443,43 @@ server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, } } - if (volinfo->backend == GD_VOL_BK_BD) { - xl = volgen_graph_add (graph, "storage/bd_map", volname); - if (!xl) - return -1; + xl = volgen_graph_add (graph, "storage/posix", volname); + if (!xl) + return -1; - ret = xlator_set_option (xl, "device", "vg"); - if (ret) - return -1; + ret = xlator_set_option (xl, "directory", path); + if (ret) + return -1; - vg = gf_strdup (path); - vgname = strrchr (vg, '/'); - if (strchr(vg, '/') != vgname) { - gf_log ("glusterd", GF_LOG_ERROR, - "invalid vg specified %s", path); - GF_FREE (vg); - goto out; - } - vgname++; - ret = xlator_set_option (xl, "export", vgname); - GF_FREE (vg); - if (ret) - return -1; - } else { - xl = volgen_graph_add (graph, "storage/posix", volname); + ret = xlator_set_option (xl, "volume-id", + uuid_utoa (volinfo->volume_id)); + if (ret) + return -1; + + ret = check_and_add_debug_xl (graph, set_dict, volname, + "posix"); + if (ret) + return -1; +#ifdef HAVE_BD_XLATOR + if (*brickinfo->vg != '\0') { + /* Now add BD v2 xlator if volume is BD type */ + xl = volgen_graph_add (graph, "storage/bd", volname); if (!xl) return -1; - ret = xlator_set_option (xl, "directory", path); + ret = xlator_set_option (xl, "device", "vg"); if (ret) return -1; - - ret = xlator_set_option (xl, "volume-id", - uuid_utoa (volinfo->volume_id)); + ret = xlator_set_option (xl, "export", brickinfo->vg); if (ret) return -1; - ret = check_and_add_debug_xl (graph, set_dict, volname, - "posix"); + ret = check_and_add_debug_xl (graph, set_dict, volname, "bd"); if (ret) return -1; + } +#endif xl = volgen_graph_add (graph, "features/changelog", volname); if (!xl) @@ -1616,7 +1614,8 @@ server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, } /* Check for read-only volume option, and add it to the graph */ - if (dict_get_str_boolean (set_dict, "features.read-only", 0)) { + if (dict_get_str_boolean (set_dict, "features.read-only", 0) + || volinfo -> is_snap_volume) { xl = volgen_graph_add (graph, "features/read-only", volname); if (!xl) { ret = -1; @@ -1633,6 +1632,18 @@ server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, } } + /* Check for compress volume option, and add it to the graph on server side */ + if (dict_get_str_boolean (set_dict, "features.compress", 0)) { + xl = volgen_graph_add (graph, "features/cdc", volname); + if (!xl) { + ret = -1; + goto out; + } + ret = dict_set_str (set_dict, "compress.mode", "server"); + if (ret) + goto out; + } + xl = volgen_graph_add_as (graph, "debug/io-stats", path); if (!xl) return -1; @@ -2444,6 +2455,29 @@ out: return ret; } +static int client_graph_set_perf_options(volgen_graph_t *graph, + glusterd_volinfo_t *volinfo, + dict_t *set_dict) +{ + data_t *tmp_data = NULL; + char *volname = NULL; + + /* + * Logic to make sure NFS doesn't have performance translators by + * default for a volume + */ + volname = volinfo->volname; + tmp_data = dict_get (set_dict, "nfs-volume-file"); + if (!tmp_data) + return volgen_graph_set_options_generic(graph, set_dict, + volname, + &perfxl_option_handler); + else + return volgen_graph_set_options_generic(graph, set_dict, + volname, + &nfsperfxl_option_handler); +} + static int client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, dict_t *set_dict, void *param) @@ -2451,7 +2485,6 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, int ret = 0; xlator_t *xl = NULL; char *volname = NULL; - data_t *tmp_data = NULL; volname = volinfo->volname; ret = volgen_graph_build_clients (graph, volinfo, set_dict, param); @@ -2462,6 +2495,31 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, if (ret == -1) goto out; + /* Check for compress volume option, and add it to the graph on client side */ + if (dict_get_str_boolean (set_dict, "features.compress", 0)) { + xl = volgen_graph_add (graph, "features/cdc", volname); + if (!xl) { + ret = -1; + goto out; + } + ret = dict_set_str (set_dict, "compress.mode", "client"); + if (ret) + goto out; + + } + + ret = glusterd_volinfo_get_boolean (volinfo, "features.encryption"); + if (ret == -1) + goto out; + if (ret) { + xl = volgen_graph_add (graph, "encryption/crypt", volname); + + if (!xl) { + ret = -1; + goto out; + } + } + ret = glusterd_volinfo_get_boolean (volinfo, VKEY_FEATURES_QUOTA); if (ret == -1) goto out; @@ -2474,16 +2532,20 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, } } - /* Logic to make sure NFS doesn't have performance translators by - default for a volume */ - tmp_data = dict_get (set_dict, "nfs-volume-file"); - if (!tmp_data) - ret = volgen_graph_set_options_generic (graph, set_dict, volinfo, - &perfxl_option_handler); - else - ret = volgen_graph_set_options_generic (graph, set_dict, volname, - &nfsperfxl_option_handler); + ret = glusterd_volinfo_get_boolean (volinfo, "features.file-snapshot"); + if (ret == -1) + goto out; + if (ret) { + xl = volgen_graph_add (graph, "features/qemu-block", volname); + + if (!xl) { + ret = -1; + goto out; + } + } + + ret = client_graph_set_perf_options(graph, volinfo, set_dict); if (ret) goto out; @@ -3113,8 +3175,6 @@ glusterd_generate_brick_volfile (glusterd_volinfo_t *volinfo, return ret; } - - static void get_vol_tstamp_file (char *filename, glusterd_volinfo_t *volinfo) { @@ -3215,7 +3275,7 @@ enumerate_transport_reqs (gf_transport_type type, char **types) } } -static int +int generate_client_volfiles (glusterd_volinfo_t *volinfo, glusterd_client_type_t client_type) { @@ -3381,6 +3441,54 @@ out: } int +glusterd_check_nfs_topology_identical (gf_boolean_t *identical) +{ + char nfsvol[PATH_MAX] = {0,}; + char tmpnfsvol[PATH_MAX] = {0,}; + glusterd_conf_t *conf = NULL; + xlator_t *this = THIS; + int ret = -1; + int tmpclean = 0; + int tmpfd = -1; + + if ((!identical) || (!this) || (!this->private)) + goto out; + + conf = (glusterd_conf_t *) this->private; + + /* Fetch the original NFS volfile */ + glusterd_get_nodesvc_volfile ("nfs", conf->workdir, + nfsvol, sizeof (nfsvol)); + + /* Create the temporary NFS volfile */ + snprintf (tmpnfsvol, sizeof (tmpnfsvol), "/tmp/gnfs-XXXXXX"); + tmpfd = mkstemp (tmpnfsvol); + if (tmpfd < 0) { + gf_log (this->name, GF_LOG_WARNING, + "Unable to create temp file %s: (%s)", + tmpnfsvol, strerror (errno)); + goto out; + } + + tmpclean = 1; /* SET the flag to unlink() tmpfile */ + + ret = glusterd_create_global_volfile (build_nfs_graph, + tmpnfsvol, NULL); + if (ret) + goto out; + + /* Compare the topology of volfiles */ + ret = glusterd_check_topology_identical (nfsvol, tmpnfsvol, + identical); +out: + if (tmpfd >= 0) + close (tmpfd); + if (tmpclean) + unlink (tmpnfsvol); + return ret; +} + +int glusterd_check_nfs_volfile_identical (gf_boolean_t *identical) { char nfsvol[PATH_MAX] = {0,}; @@ -3829,3 +3937,129 @@ gd_is_boolean_option (char *key) return _gf_false; } + +/* This function will restore origin volume to it's snap. + * The restore operation will simply replace the Gluster origin + * volume with the snap volume. + * TODO: Multi-volume delete to be done. + * Cleanup in case of restore failure is pending. + * + * @param orig_vol volinfo of origin volume + * @param snap_vol volinfo of snapshot volume + * + * @return 0 on success and negative value on error + */ +int +gd_restore_snap_volume (dict_t *rsp_dict, + glusterd_volinfo_t *orig_vol, + glusterd_volinfo_t *snap_vol) +{ + int ret = -1; + glusterd_volinfo_t *new_volinfo = NULL; + glusterd_snap_t *snap = NULL; + xlator_t *this = NULL; + glusterd_conf_t *conf = NULL; + glusterd_volinfo_t *temp_volinfo = NULL; + glusterd_volinfo_t *voliter = NULL; + + this = THIS; + GF_ASSERT (this); + GF_ASSERT (rsp_dict); + conf = this->private; + GF_ASSERT (conf); + + GF_VALIDATE_OR_GOTO (this->name, orig_vol, out); + GF_VALIDATE_OR_GOTO (this->name, snap_vol, out); + snap = snap_vol->snapshot; + GF_VALIDATE_OR_GOTO (this->name, snap, out); + + /* Snap volume must be stoped before performing the + * restore operation. + */ + ret = glusterd_stop_volume (snap_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to stop " + "snap volume"); + goto out; + } + + /* Create a new volinfo for the restored volume */ + ret = glusterd_volinfo_dup (snap_vol, &new_volinfo, _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to create volinfo"); + goto out; + } + + /* Following entries need to be derived from origin volume. */ + strcpy (new_volinfo->volname, orig_vol->volname); + uuid_copy (new_volinfo->volume_id, orig_vol->volume_id); + new_volinfo->snap_count = orig_vol->snap_count; + new_volinfo->snap_max_hard_limit = orig_vol->snap_max_hard_limit; + new_volinfo->is_volume_restored = _gf_true; + + /* Bump the version of the restored volume, so that nodes * + * which are done can sync during handshake */ + new_volinfo->version = orig_vol->version; + + list_for_each_entry_safe (voliter, temp_volinfo, + &orig_vol->snap_volumes, snapvol_list) { + list_add_tail (&voliter->snapvol_list, + &new_volinfo->snap_volumes); + } + /* Copy the snap vol info to the new_volinfo.*/ + ret = glusterd_snap_volinfo_restore (rsp_dict, new_volinfo, snap_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to restore snap"); + (void)glusterd_volinfo_delete (new_volinfo); + goto out; + } + + /* If the orig_vol is already restored then we should delete + * the backend LVMs */ + if (orig_vol->is_volume_restored) { + ret = glusterd_lvm_snapshot_remove (rsp_dict, orig_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to remove " + "LVM backend"); + (void)glusterd_volinfo_delete (new_volinfo); + goto out; + } + } + + /* Once the new_volinfo is completely constructed then delete + * the orinal volinfo + */ + ret = glusterd_volinfo_delete (orig_vol); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to delete volinfo"); + (void)glusterd_volinfo_delete (new_volinfo); + goto out; + } + + /* New volinfo always shows the status as created. Therefore + * set the status to stop. */ + glusterd_set_volume_status (new_volinfo, GLUSTERD_STATUS_STOPPED); + + list_add_tail (&new_volinfo->vol_list, &conf->volumes); + + /* Now delete the snap entry. As a first step delete the snap + * volume information stored in store. */ + ret = glusterd_snap_remove (rsp_dict, snap, _gf_false, _gf_true); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, "Failed to delete " + "snap %s", snap->snapname); + goto out; + } + + ret = glusterd_store_volinfo (new_volinfo, + GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to store volinfo"); + goto out; + } + + ret = 0; +out: + + return ret; +} diff --git a/xlators/mgmt/glusterd/src/glusterd-volgen.h b/xlators/mgmt/glusterd/src/glusterd-volgen.h index 4ff899f4a..fcbaaf93e 100644 --- a/xlators/mgmt/glusterd/src/glusterd-volgen.h +++ b/xlators/mgmt/glusterd/src/glusterd-volgen.h @@ -23,6 +23,7 @@ #define VKEY_DIAG_LAT_MEASUREMENT "diagnostics.latency-measurement" #define VKEY_FEATURES_LIMIT_USAGE "features.limit-usage" #define VKEY_MARKER_XTIME GEOREP".indexing" +#define VKEY_MARKER_XTIME_FORCE GEOREP".ignore-pid-check" #define VKEY_CHANGELOG "changelog.changelog" #define VKEY_FEATURES_QUOTA "features.quota" @@ -74,6 +75,7 @@ typedef enum { GF_XLATOR_INDEX, GF_XLATOR_MARKER, GF_XLATOR_IO_STATS, + GF_XLATOR_BD, GF_XLATOR_NONE, } glusterd_server_xlator_t; @@ -122,6 +124,10 @@ int glusterd_create_shd_volfile (); int glusterd_delete_volfile (glusterd_volinfo_t *volinfo, glusterd_brickinfo_t *brickinfo); +int +glusterd_delete_snap_volfile (glusterd_volinfo_t *volinfo, + glusterd_volinfo_t *snap_volinfo, + glusterd_brickinfo_t *brickinfo); int glusterd_volinfo_get (glusterd_volinfo_t *volinfo, char *key, char **value); int glusterd_volinfo_get_boolean (glusterd_volinfo_t *volinfo, char *key); @@ -135,11 +141,22 @@ glusterd_check_voloption_flags (char *key, int32_t flags); gf_boolean_t glusterd_is_valid_volfpath (char *volname, char *brick); int generate_brick_volfiles (glusterd_volinfo_t *volinfo); +int generate_snap_brick_volfiles (glusterd_volinfo_t *volinfo, + glusterd_volinfo_t *snap_volinfo); +int generate_client_volfiles (glusterd_volinfo_t *volinfo, + glusterd_client_type_t client_type); +int +generate_snap_client_volfiles (glusterd_volinfo_t *actual_volinfo, + glusterd_volinfo_t *snap_volinfo, + glusterd_client_type_t client_type, + gf_boolean_t vol_restore); int glusterd_get_volopt_content (dict_t *dict, gf_boolean_t xml_out); char* glusterd_get_trans_type_rb (gf_transport_type ttype); int glusterd_check_nfs_volfile_identical (gf_boolean_t *identical); +int +glusterd_check_nfs_topology_identical (gf_boolean_t *identical); uint32_t glusterd_get_op_version_for_key (char *key); @@ -153,4 +170,7 @@ gd_is_xlator_option (char *key); gf_boolean_t gd_is_boolean_option (char *key); +int gd_restore_snap_volume (dict_t *rsp_dict, + glusterd_volinfo_t *orig_vol, + glusterd_volinfo_t *snap_vol); #endif diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c index c1aea4136..0d322b9ad 100644 --- a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c +++ b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c @@ -30,6 +30,7 @@ #define glusterd_op_start_volume_args_get(dict, volname, flags) \ glusterd_op_stop_volume_args_get (dict, volname, flags) + int __glusterd_handle_create_volume (rpcsvc_request_t *req) { @@ -604,31 +605,96 @@ glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req) } #ifdef HAVE_BD_XLATOR +/* + * Validates if given VG in the brick exists or not. Also checks if VG has + * GF_XATTR_VOL_ID_KEY tag set to avoid using same VG for multiple bricks. + * Tag is checked only during glusterd_op_stage_create_volume. Tag is set during + * glusterd_validate_and_create_brickpath(). + * @brick - brick info, @check_tag - check for VG tag or not + * @msg - Error message to return to caller + */ int -glusterd_is_valid_vg (const char *name) +glusterd_is_valid_vg (glusterd_brickinfo_t *brick, int check_tag, char *msg) { - lvm_t handle = NULL; - vg_t vg = NULL; - char *vg_name = NULL; - int retval = -1; + lvm_t handle = NULL; + vg_t vg = NULL; + char *vg_name = NULL; + int retval = 0; + char *p = NULL; + char *ptr = NULL; + struct dm_list *dm_lvlist = NULL; + struct dm_list *dm_seglist = NULL; + struct lvm_lv_list *lv_list = NULL; + struct lvm_property_value prop = {0, }; + struct lvm_lvseg_list *seglist = NULL; + struct dm_list *taglist = NULL; + struct lvm_str_list *strl = NULL; handle = lvm_init (NULL); if (!handle) { - gf_log ("", GF_LOG_ERROR, "lvm_init failed"); + sprintf (msg, "lvm_init failed, could not validate vg"); return -1; } - vg_name = gf_strdup (name); - vg = lvm_vg_open (handle, basename (vg_name), "r", 0); + if (*brick->vg == '\0') { /* BD xlator has vg in brick->path */ + p = gf_strdup (brick->path); + vg_name = strtok_r (p, "/", &ptr); + } else + vg_name = brick->vg; + + vg = lvm_vg_open (handle, vg_name, "r", 0); if (!vg) { - gf_log ("", GF_LOG_ERROR, "no such vg: %s", vg_name); + sprintf (msg, "no such vg: %s", vg_name); + retval = -1; goto out; } + if (!check_tag) + goto next; + + taglist = lvm_vg_get_tags (vg); + if (!taglist) + goto next; + + dm_list_iterate_items (strl, taglist) { + if (!strncmp(strl->str, GF_XATTR_VOL_ID_KEY, + strlen (GF_XATTR_VOL_ID_KEY))) { + sprintf (msg, "VG %s is already part of" + " a brick", vg_name); + retval = -1; + goto out; + } + } +next: + + brick->caps = CAPS_BD | CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT; + + dm_lvlist = lvm_vg_list_lvs (vg); + if (!dm_lvlist) + goto out; + + dm_list_iterate_items (lv_list, dm_lvlist) { + dm_seglist = lvm_lv_list_lvsegs (lv_list->lv); + dm_list_iterate_items (seglist, dm_seglist) { + prop = lvm_lvseg_get_property (seglist->lvseg, + "segtype"); + if (!prop.is_valid || !prop.value.string) + continue; + if (!strcmp (prop.value.string, "thin-pool")) { + brick->caps |= CAPS_THIN; + gf_log (THIS->name, GF_LOG_INFO, "Thin Pool " + "\"%s\" will be used for thin LVs", + lvm_lv_get_name (lv_list->lv)); + break; + } + } + } + retval = 0; out: if (vg) lvm_vg_close (vg); lvm_quit (handle); - GF_FREE (vg_name); + if (p) + GF_FREE (p); return retval; } #endif @@ -653,9 +719,6 @@ glusterd_op_stage_create_volume (dict_t *dict, char **op_errstr) char msg[2048] = {0}; uuid_t volume_uuid; char *volume_uuid_str; -#ifdef HAVE_BD_XLATOR - char *dev_type = NULL; -#endif gf_boolean_t is_force = _gf_false; this = THIS; @@ -700,10 +763,6 @@ glusterd_op_stage_create_volume (dict_t *dict, char **op_errstr) goto out; } -#ifdef HAVE_BD_XLATOR - ret = dict_get_str (dict, "device", &dev_type); -#endif - ret = dict_get_str (dict, "bricks", &bricks); if (ret) { gf_log (this->name, GF_LOG_ERROR, "Unable to get bricks for " @@ -752,19 +811,15 @@ glusterd_op_stage_create_volume (dict_t *dict, char **op_errstr) goto out; } + if (!uuid_compare (brick_info->uuid, MY_UUID)) { + #ifdef HAVE_BD_XLATOR - if (dev_type) { - ret = glusterd_is_valid_vg (brick_info->path); - if (ret) { - snprintf (msg, sizeof(msg), "invalid vg %s", - brick_info->path); - goto out; + if (brick_info->vg[0]) { + ret = glusterd_is_valid_vg (brick_info, 1, msg); + if (ret) + goto out; } - - break; - } else #endif - if (!uuid_compare (brick_info->uuid, MY_UUID)) { ret = glusterd_validate_and_create_brickpath (brick_info, volume_uuid, op_errstr, is_force); @@ -862,6 +917,7 @@ glusterd_op_stage_start_volume (dict_t *dict, char **op_errstr) uuid_t volume_id = {0,}; char volid[50] = {0,}; char xattr_volid[50] = {0,}; + int caps = 0; this = THIS; GF_ASSERT (this); @@ -911,9 +967,6 @@ glusterd_op_stage_start_volume (dict_t *dict, char **op_errstr) if (uuid_compare (brickinfo->uuid, MY_UUID)) continue; - if (volinfo->backend == GD_VOL_BK_BD) - continue; - ret = gf_lstat_dir (brickinfo->path, NULL); if (ret && (flags & GF_CLI_FLAG_OP_FORCE)) { continue; @@ -958,8 +1011,24 @@ glusterd_op_stage_start_volume (dict_t *dict, char **op_errstr) ret = -1; goto out; } +#ifdef HAVE_BD_XLATOR + if (brickinfo->vg[0]) + caps = CAPS_BD | CAPS_THIN | + CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT; + /* Check for VG/thin pool if its BD volume */ + if (brickinfo->vg[0]) { + ret = glusterd_is_valid_vg (brickinfo, 0, msg); + if (ret) + goto out; + /* if anyone of the brick does not have thin support, + disable it for entire volume */ + caps &= brickinfo->caps; + } else + caps = 0; +#endif } + volinfo->caps = caps; ret = 0; out: if (ret && (msg[0] != '\0')) { @@ -1114,6 +1183,16 @@ glusterd_op_stage_delete_volume (dict_t *dict, char **op_errstr) goto out; } + if (volinfo->snap_count > 0 || !list_empty(&volinfo->snap_volumes)) { + snprintf (msg, sizeof (msg), "Cannot delete Volume %s ," + "as it has %ld snapshots. " + "To delete the volume, " + "first delete all the snapshots under it.", + volname, volinfo->snap_count); + ret = -1; + goto out; + } + ret = 0; out: @@ -1211,14 +1290,22 @@ glusterd_op_stage_heal_volume (dict_t *dict, char **op_errstr) goto out; } - if ((heal_op != GF_AFR_OP_INDEX_SUMMARY) && - !glusterd_is_nodesvc_online ("glustershd")) { - ret = -1; - *op_errstr = gf_strdup ("Self-heal daemon is not running." - " Check self-heal daemon log file."); - gf_log (this->name, GF_LOG_WARNING, "%s", "Self-heal daemon is " - "not running. Check self-heal daemon log file."); - goto out; + switch (heal_op) { + case GF_AFR_OP_INDEX_SUMMARY: + case GF_AFR_OP_STATISTICS_HEAL_COUNT: + case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA: + break; + default: + if (!glusterd_is_nodesvc_online("glustershd")){ + ret = -1; + *op_errstr = gf_strdup ("Self-heal daemon is " + "not running. Check self-heal " + "daemon log file."); + gf_log (this->name, GF_LOG_WARNING, "%s", + "Self-heal daemon is not running." + "Check self-heal daemon log file."); + goto out; + } } ret = 0; @@ -1342,109 +1429,6 @@ out: return ret; } -#ifdef HAVE_BD_XLATOR -int -glusterd_op_stage_bd (dict_t *dict, char **op_errstr) -{ - int ret = -1; - char *volname = NULL; - char *path = NULL; - char *size = NULL; - glusterd_volinfo_t *volinfo = NULL; - char msg[2048] = {0,}; - gf_xl_bd_op_t bd_op = GF_BD_OP_INVALID; - uint64_t bytes = 0; - - ret = dict_get_str (dict, "volname", &volname); - if (ret) { - snprintf (msg, sizeof(msg), "Failed to get volume name"); - gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - goto out; - } - - ret = dict_get_int32 (dict, "bd-op", (int32_t *)&bd_op); - if (ret) { - snprintf (msg, sizeof(msg), "Failed to get bd-op"); - gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - goto out; - } - - ret = dict_get_str (dict, "path", &path); - if (ret) { - snprintf (msg, sizeof(msg), "Failed to get path"); - gf_log (THIS->name, GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - goto out; - } - - if (bd_op == GF_BD_OP_NEW_BD) { - ret = dict_get_str (dict, "size", &size); - if (ret) { - snprintf (msg, sizeof(msg), "Failed to get size"); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - goto out; - } - if (gf_string2bytesize (size, &bytes) < 0) { - snprintf (msg, sizeof(msg), - "Invalid size %s, suffix with KB, MB etc", - size); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - ret = -1; - goto out; - } - } else if (bd_op == GF_BD_OP_SNAPSHOT_BD) { - ret = dict_get_str (dict, "size", &size); - if (ret) { - snprintf (msg, sizeof(msg), "Failed to get size"); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - goto out; - } - - if (gf_string2bytesize (size, &bytes) < 0) { - ret = -1; - snprintf (msg, sizeof(msg), - "Invalid size %s, suffix with KB, MB etc", - size); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - goto out; - } - } - - ret = glusterd_volinfo_find (volname, &volinfo); - if (ret) { - snprintf (msg, sizeof(msg), "Volume %s does not exist", - volname); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - goto out; - } - - ret = glusterd_validate_volume_id (dict, volinfo); - if (ret) - goto out; - - if (!glusterd_is_volume_started (volinfo)) { - snprintf (msg, sizeof(msg), "Volume %s is not started", - volname); - gf_log ("", GF_LOG_ERROR, "%s", msg); - *op_errstr = gf_strdup (msg); - ret = -1; - goto out; - } - - ret = 0; -out: - gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); - return ret; -} -#endif - int glusterd_op_create_volume (dict_t *dict, char **op_errstr) { @@ -1466,9 +1450,8 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) char *str = NULL; char *username = NULL; char *password = NULL; -#ifdef HAVE_BD_XLATOR - char *device = NULL; -#endif + int caps = 0; + char msg[1024] __attribute__((unused)) = {0, }; this = THIS; GF_ASSERT (this); @@ -1523,12 +1506,6 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) goto out; } -#ifdef HAVE_BD_XLATOR - ret = dict_get_str (dict, "device", &device); - if (!ret) - volinfo->backend = GD_VOL_BK_BD; -#endif - /* replica-count 1 means, no replication, file is in one brick only */ volinfo->replica_count = 1; /* stripe-count 1 means, no striping, file is present as a whole */ @@ -1637,6 +1614,7 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) if (count) brick = strtok_r (brick_list+1, " \n", &saveptr); + caps = CAPS_BD | CAPS_THIN | CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT; while ( i <= count) { ret = glusterd_brickinfo_new_from_brick (brick, &brickinfo); @@ -1649,6 +1627,27 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) brickinfo->hostname, brickinfo->path); goto out; } + +#ifdef HAVE_BD_XLATOR + if (!uuid_compare (brickinfo->uuid, MY_UUID)) { + if (brickinfo->vg[0]) { + ret = glusterd_is_valid_vg (brickinfo, 0, msg); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "%s", + msg); + goto out; + } + + /* if anyone of the brick does not have thin + support, disable it for entire volume */ + caps &= brickinfo->caps; + + + } else + caps = 0; + } +#endif + list_add_tail (&brickinfo->brick_list, &volinfo->bricks); brick = strtok_r (NULL, " \n", &saveptr); i++; @@ -1656,6 +1655,8 @@ glusterd_op_create_volume (dict_t *dict, char **op_errstr) gd_update_volume_op_versions (volinfo); + volinfo->caps = caps; + ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); if (ret) { glusterd_store_delete_volume (volinfo); @@ -1726,6 +1727,47 @@ out: return ret; } +int +glusterd_stop_volume (glusterd_volinfo_t *volinfo) +{ + int ret = -1; + glusterd_brickinfo_t *brickinfo = NULL; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + GF_VALIDATE_OR_GOTO (this->name, volinfo, out); + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + ret = glusterd_brick_stop (volinfo, brickinfo, _gf_false); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to stop " + "brick (%s)", brickinfo->path); + goto out; + } + } + + glusterd_set_volume_status (volinfo, GLUSTERD_STATUS_STOPPED); + + ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to store volinfo of " + "%s volume", volinfo->volname); + goto out; + } + + ret = glusterd_nodesvcs_handle_graph_change (volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to notify graph " + "change for %s volume", volinfo->volname); + goto out; + } + +out: + return ret; +} + int glusterd_op_stop_volume (dict_t *dict) @@ -1734,7 +1776,6 @@ glusterd_op_stop_volume (dict_t *dict) int flags = 0; char *volname = NULL; glusterd_volinfo_t *volinfo = NULL; - glusterd_brickinfo_t *brickinfo = NULL; xlator_t *this = NULL; this = THIS; @@ -1751,19 +1792,12 @@ glusterd_op_stop_volume (dict_t *dict) goto out; } - list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { - ret = glusterd_brick_stop (volinfo, brickinfo, _gf_false); - if (ret) - goto out; - } - - glusterd_set_volume_status (volinfo, GLUSTERD_STATUS_STOPPED); - - ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); - if (ret) + ret = glusterd_stop_volume (volinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to stop %s volume", + volname); goto out; - - ret = glusterd_nodesvcs_handle_graph_change (volinfo); + } out: return ret; } diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-set.c b/xlators/mgmt/glusterd/src/glusterd-volume-set.c index ed5fc96ad..665a8b298 100644 --- a/xlators/mgmt/glusterd/src/glusterd-volume-set.c +++ b/xlators/mgmt/glusterd/src/glusterd-volume-set.c @@ -724,6 +724,34 @@ struct volopt_map_entry glusterd_volopt_map[] = { .flags = OPT_FLAG_CLIENT_OPT }, + /* Crypt xlator options */ + + { .key = "features.encryption", + .voltype = "encryption/crypt", + .option = "!feat", + .value = "off", + .op_version = 3, + .description = "enable/disable client-side encryption for " + "the volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + + { .key = "encryption.master-key", + .voltype = "encryption/crypt", + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "encryption.data-key-size", + .voltype = "encryption/crypt", + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + { .key = "encryption.block-size", + .voltype = "encryption/crypt", + .op_version = 3, + .flags = OPT_FLAG_CLIENT_OPT + }, + /* Client xlator options */ { .key = "network.frame-timeout", .voltype = "protocol/client", @@ -808,6 +836,12 @@ struct volopt_map_entry glusterd_volopt_map[] = { .option = "statedump-path", .op_version = 1 }, + { .key = "server.outstanding-rpc-limit", + .voltype = "protocol/server", + .option = "rpc.outstanding-rpc-limit", + .type = GLOBAL_DOC, + .op_version = 3 + }, { .key = "features.lock-heal", .voltype = "protocol/server", .option = "lk-heal", @@ -845,6 +879,15 @@ struct volopt_map_entry glusterd_volopt_map[] = { .description = "enable/disable read-ahead translator in the volume.", .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT }, + { .key = "performance.readdir-ahead", + .voltype = "performance/readdir-ahead", + .option = "!perf", + .value = "off", + .op_version = 3, + .description = "enable/disable readdir-ahead translator in the volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + { .key = "performance.io-cache", .voltype = "performance/io-cache", .option = "!perf", @@ -944,6 +987,61 @@ struct volopt_map_entry glusterd_volopt_map[] = { .flags = OPT_FLAG_CLIENT_OPT }, + /* Feature translators */ + { .key = "features.file-snapshot", + .voltype = "features/qemu-block", + .option = "!feat", + .value = "off", + .op_version = 3, + .description = "enable/disable file-snapshot feature in the " + "volume.", + .flags = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT + }, + +#ifdef HAVE_LIB_Z + /* Compressor-decompressor xlator options + * defaults used from xlator/feature/compress/src/cdc.h + */ + { .key = "features.compress", + .voltype = "features/cdc", + .option = "!compress", + .value = "off", + .type = NO_DOC, + .op_version = 2, + .description = "enable/disable compression translator" + }, + { .key = "compress.mode", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "compress.window-size", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "compress.mem-level", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "compress.min-size", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "compress.compression-level", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + { .key = "compress.debug", + .voltype = "features/cdc", + .type = NO_DOC, + .op_version = 2 + }, + #endif + /* Quota xlator options */ { .key = VKEY_FEATURES_LIMIT_USAGE, .voltype = "features/quota", @@ -987,6 +1085,22 @@ struct volopt_map_entry glusterd_volopt_map[] = { .flags = OPT_FLAG_FORCE, .op_version = 1 }, + { .key = VKEY_MARKER_XTIME_FORCE, + .voltype = "features/marker", + .option = "gsync-force-xtime", + .value = "off", + .type = NO_DOC, + .flags = OPT_FLAG_FORCE, + .op_version = 2 + }, + { .key = VKEY_MARKER_XTIME_FORCE, + .voltype = "features/marker", + .option = "!gsync-force-xtime", + .value = "off", + .type = NO_DOC, + .flags = OPT_FLAG_FORCE, + .op_version = 2 + }, { .key = VKEY_FEATURES_QUOTA, .voltype = "features/marker", .option = "quota", @@ -1106,6 +1220,12 @@ struct volopt_map_entry glusterd_volopt_map[] = { .type = GLOBAL_DOC, .op_version = 1 }, + { .key = "nfs.outstanding-rpc-limit", + .voltype = "nfs/server", + .option = "rpc.outstanding-rpc-limit", + .type = GLOBAL_DOC, + .op_version = 3 + }, { .key = "nfs.port", .voltype = "nfs/server", .option = "nfs.port", @@ -1212,6 +1332,24 @@ struct volopt_map_entry glusterd_volopt_map[] = { .type = GLOBAL_DOC, .op_version = 3 }, + { .key = "nfs.read-size", + .voltype = "nfs/server", + .option = "nfs3.read-size", + .type = GLOBAL_DOC, + .op_version = 3 + }, + { .key = "nfs.write-size", + .voltype = "nfs/server", + .option = "nfs3.write-size", + .type = GLOBAL_DOC, + .op_version = 3 + }, + { .key = "nfs.readdir-size", + .voltype = "nfs/server", + .option = "nfs3.readdir-size", + .type = GLOBAL_DOC, + .op_version = 3 + }, /* Other options which don't fit any place above */ { .key = "features.read-only", @@ -1258,6 +1396,10 @@ struct volopt_map_entry glusterd_volopt_map[] = { .voltype = "storage/posix", .op_version = 3 }, + { .key = "storage.bd-aio", + .voltype = "storage/bd", + .op_version = 3 + }, { .key = "config.memory-accounting", .voltype = "configuration", .option = "!config", diff --git a/xlators/mgmt/glusterd/src/glusterd.c b/xlators/mgmt/glusterd/src/glusterd.c index a43c8d542..59288ada0 100644 --- a/xlators/mgmt/glusterd/src/glusterd.c +++ b/xlators/mgmt/glusterd/src/glusterd.c @@ -1,5 +1,5 @@ /* - Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + Copyright (c) 2006-2013 Red Hat, Inc. <http://www.redhat.com> This file is part of GlusterFS. This file is licensed to you under your choice of the GNU Lesser @@ -36,6 +36,7 @@ #include "glusterd-store.h" #include "glusterd-hooks.h" #include "glusterd-utils.h" +#include "glusterd-locks.h" #include "common-utils.h" #include "run.h" @@ -44,29 +45,43 @@ #include "glusterd-mountbroker.h" extern struct rpcsvc_program gluster_handshake_prog; +extern struct rpcsvc_program gluster_cli_getspec_prog; extern struct rpcsvc_program gluster_pmap_prog; extern glusterd_op_info_t opinfo; extern struct rpcsvc_program gd_svc_mgmt_prog; +extern struct rpcsvc_program gd_svc_mgmt_v3_prog; extern struct rpcsvc_program gd_svc_peer_prog; extern struct rpcsvc_program gd_svc_cli_prog; +extern struct rpcsvc_program gd_svc_cli_prog_ro; extern struct rpc_clnt_program gd_brick_prog; extern struct rpcsvc_program glusterd_mgmt_hndsk_prog; +extern char snap_mount_folder[PATH_MAX]; + rpcsvc_cbk_program_t glusterd_cbk_prog = { .progname = "Gluster Callback", .prognum = GLUSTER_CBK_PROGRAM, .progver = GLUSTER_CBK_VERSION, }; -struct rpcsvc_program *all_programs[] = { +struct rpcsvc_program *gd_inet_programs[] = { &gd_svc_peer_prog, - &gd_svc_cli_prog, + &gd_svc_cli_prog_ro, &gd_svc_mgmt_prog, + &gd_svc_mgmt_v3_prog, &gluster_pmap_prog, &gluster_handshake_prog, &glusterd_mgmt_hndsk_prog, }; -int rpcsvc_programs_count = (sizeof (all_programs) / sizeof (all_programs[0])); +int gd_inet_programs_count = (sizeof (gd_inet_programs) / + sizeof (gd_inet_programs[0])); + +struct rpcsvc_program *gd_uds_programs[] = { + &gd_svc_cli_prog, + &gluster_cli_getspec_prog, +}; +int gd_uds_programs_count = (sizeof (gd_uds_programs) / + sizeof (gd_uds_programs[0])); const char *gd_op_list[GD_OP_MAX + 1] = { [GD_OP_NONE] = "Invalid op", @@ -97,6 +112,7 @@ const char *gd_op_list[GD_OP_MAX + 1] = { [GD_OP_COPY_FILE] = "Copy File", [GD_OP_SYS_EXEC] = "Execute system commands", [GD_OP_GSYNC_CREATE] = "Geo-replication Create", + [GD_OP_SNAP] = "Snapshot", [GD_OP_MAX] = "Invalid op" }; @@ -122,12 +138,12 @@ glusterd_uuid_init () GF_ASSERT (this); priv = this->private; - ret = glusterd_retrieve_uuid (); - if (ret == 0) { - gf_log (this->name, GF_LOG_INFO, - "retrieved UUID: %s", uuid_utoa (priv->uuid)); - return 0; - } + ret = glusterd_retrieve_uuid (); + if (ret == 0) { + gf_log (this->name, GF_LOG_INFO, + "retrieved UUID: %s", uuid_utoa (priv->uuid)); + return 0; + } ret = glusterd_uuid_generate_save (); @@ -601,21 +617,21 @@ configure_syncdaemon (glusterd_conf_t *conf) /* pid-file */ runinit_gsyncd_setrx (&runner, conf); runner_add_arg (&runner, "pid-file"); - runner_argprintf (&runner, "%s/${mastervol}-${slavevol}/${eSlave}.pid", georepdir); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.pid", georepdir); runner_add_args (&runner, ".", ".", NULL); RUN_GSYNCD_CMD; /* state-file */ runinit_gsyncd_setrx (&runner, conf); runner_add_arg (&runner, "state-file"); - runner_argprintf (&runner, "%s/${mastervol}-${slavevol}/${eSlave}.status", georepdir); + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.status", georepdir); runner_add_args (&runner, ".", ".", NULL); RUN_GSYNCD_CMD; /* state-detail-file */ runinit_gsyncd_setrx (&runner, conf); runner_add_arg (&runner, "state-detail-file"); - runner_argprintf (&runner, "%s/${mastervol}-${slavevol}/${eSlave}-detail.status", + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}-detail.status", georepdir); runner_add_args (&runner, ".", ".", NULL); RUN_GSYNCD_CMD; @@ -623,7 +639,7 @@ configure_syncdaemon (glusterd_conf_t *conf) /* state-detail-file */ runinit_gsyncd_setrx (&runner, conf); runner_add_arg (&runner, "state-detail-file"); - runner_argprintf (&runner, "%s/${mastervol}-${slavevol}/${eSlave}-detail.status", + runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}-detail.status", georepdir); runner_add_args (&runner, ".", ".", NULL); RUN_GSYNCD_CMD; @@ -937,6 +953,187 @@ glusterd_launch_synctask (synctask_fn_t fn, void *opaque) " and other volume related services"); } +int +glusterd_uds_rpcsvc_notify (rpcsvc_t *rpc, void *xl, rpcsvc_event_t event, + void *data) +{ + /* glusterd_rpcsvc_notify() does stuff that calls coming in from the + * unix domain socket don't need. This is just an empty function to be + * used for the uds listener. This will be used later if required. + */ + return 0; +} + +/* The glusterd unix domain socket listener only listens for cli */ +rpcsvc_t * +glusterd_init_uds_listener (xlator_t *this) +{ + int ret = -1; + dict_t *options = NULL; + rpcsvc_t *rpc = NULL; + data_t *sock_data = NULL; + char sockfile[PATH_MAX+1] = {0,}; + int i = 0; + + + GF_ASSERT (this); + + sock_data = dict_get (this->options, "glusterd-sockfile"); + if (!sock_data) { + strncpy (sockfile, DEFAULT_GLUSTERD_SOCKFILE, PATH_MAX); + } else { + strncpy (sockfile, sock_data->data, PATH_MAX); + } + + options = dict_new (); + if (!options) + goto out; + + ret = rpcsvc_transport_unix_options_build (&options, sockfile); + if (ret) + goto out; + + rpc = rpcsvc_init (this, this->ctx, options, 8); + if (rpc == NULL) { + ret = -1; + goto out; + } + + ret = rpcsvc_register_notify (rpc, glusterd_uds_rpcsvc_notify, + this); + if (ret) { + gf_log (this->name, GF_LOG_DEBUG, + "Failed to register notify function"); + goto out; + } + + ret = rpcsvc_create_listeners (rpc, options, this->name); + if (ret != 1) { + gf_log (this->name, GF_LOG_DEBUG, "Failed to create listener"); + goto out; + } + ret = 0; + + for (i = 0; i < gd_uds_programs_count; i++) { + ret = glusterd_program_register (this, rpc, gd_uds_programs[i]); + if (ret) { + i--; + for (; i >= 0; i--) + rpcsvc_program_unregister (rpc, + gd_uds_programs[i]); + + goto out; + } + } + +out: + if (options) + dict_unref (options); + + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Failed to start glusterd " + "unix domain socket listener."); + if (rpc) { + GF_FREE (rpc); + rpc = NULL; + } + } + return rpc; +} + +void +glusterd_stop_uds_listener (xlator_t *this) +{ + glusterd_conf_t *conf = NULL; + rpcsvc_listener_t *listener = NULL; + rpcsvc_listener_t *next = NULL; + + GF_ASSERT (this); + conf = this->private; + + (void) rpcsvc_program_unregister (conf->uds_rpc, &gd_svc_cli_prog); + (void) rpcsvc_program_unregister (conf->uds_rpc, &gluster_handshake_prog); + + list_for_each_entry_safe (listener, next, &conf->uds_rpc->listeners, + list) { + rpcsvc_listener_destroy (listener); + } + + (void) rpcsvc_unregister_notify (conf->uds_rpc, glusterd_rpcsvc_notify, + this); + + unlink (DEFAULT_GLUSTERD_SOCKFILE); + + GF_FREE (conf->uds_rpc); + conf->uds_rpc = NULL; + + return; +} + +static int +glusterd_init_snap_folder (xlator_t *this) +{ + int ret = -1; + struct stat buf = {0,}; + + GF_ASSERT (this); + + /* Snapshot volumes are mounted under /var/run/gluster/snaps folder. + * But /var/run is normally a symbolic link to /run folder, which + * creates problems as the entry point in the mtab for the mount point + * and glusterd maintained entry point will be different. Therefore + * identify the correct run folder and use it for snap volume mounting. + */ + ret = lstat (GLUSTERD_VAR_RUN_DIR, &buf); + if (ret != 0) { + gf_log (this->name, GF_LOG_ERROR, + "stat fails on %s, exiting. (errno = %d)", + GLUSTERD_VAR_RUN_DIR, errno); + goto out; + } + + /* If /var/run is symlink then use /run folder */ + if (S_ISLNK (buf.st_mode)) { + strcpy (snap_mount_folder, GLUSTERD_RUN_DIR); + } else { + strcpy (snap_mount_folder, GLUSTERD_VAR_RUN_DIR); + } + + strcat (snap_mount_folder, GLUSTERD_DEFAULT_SNAPS_BRICK_DIR); + + ret = stat (snap_mount_folder, &buf); + if ((ret != 0) && (ENOENT != errno)) { + gf_log (this->name, GF_LOG_ERROR, + "stat fails on %s, exiting. (errno = %d)", + snap_mount_folder, errno); + ret = -1; + goto out; + } + + if ((!ret) && (!S_ISDIR(buf.st_mode))) { + gf_log (this->name, GF_LOG_CRITICAL, + "Provided snap path %s is not a directory," + "exiting", snap_mount_folder); + ret = -1; + goto out; + } + + if ((-1 == ret) && (ENOENT == errno)) { + /* Create missing folders */ + ret = mkdir_p (snap_mount_folder, 0777, _gf_false); + + if (-1 == ret) { + gf_log (this->name, GF_LOG_CRITICAL, + "Unable to create directory %s" + " ,errno = %d", snap_mount_folder, errno); + goto out; + } + } + +out: + return ret; +} + /* * init - called during glusterd initialization * @@ -948,6 +1145,7 @@ init (xlator_t *this) { int32_t ret = -1; rpcsvc_t *rpc = NULL; + rpcsvc_t *uds_rpc = NULL; glusterd_conf_t *conf = NULL; data_t *dir_data = NULL; struct stat buf = {0,}; @@ -973,7 +1171,7 @@ init (xlator_t *this) if ((ret != 0) && (ENOENT != errno)) { gf_log (this->name, GF_LOG_ERROR, "stat fails on %s, exiting. (errno = %d)", - workdir, errno); + workdir, errno); exit (1); } @@ -1002,6 +1200,13 @@ init (xlator_t *this) gf_log (this->name, GF_LOG_INFO, "Using %s as working directory", workdir); + ret = glusterd_init_snap_folder (this); + if (ret) { + gf_log (this->name, GF_LOG_CRITICAL, "Unable to create " + "snap backend folder"); + exit (1); + } + snprintf (cmd_log_filename, PATH_MAX,"%s/.cmd_log_history", DEFAULT_LOG_FILE_DIRECTORY); ret = gf_cmd_log_init (cmd_log_filename); @@ -1099,18 +1304,29 @@ init (xlator_t *this) goto out; } - for (i = 0; i < rpcsvc_programs_count; i++) { - ret = glusterd_program_register (this, rpc, all_programs[i]); + for (i = 0; i < gd_inet_programs_count; i++) { + ret = glusterd_program_register (this, rpc, + gd_inet_programs[i]); if (ret) { i--; for (; i >= 0; i--) rpcsvc_program_unregister (rpc, - all_programs[i]); + gd_inet_programs[i]); goto out; } } + /* Start a unix domain socket listener just for cli commands + * This should prevent ports from being wasted by being in TIMED_WAIT + * when cli commands are done continuously + */ + uds_rpc = glusterd_init_uds_listener (this); + if (uds_rpc == NULL) { + ret = -1; + goto out; + } + conf = GF_CALLOC (1, sizeof (glusterd_conf_t), gf_gld_mt_glusterd_conf_t); GF_VALIDATE_OR_GOTO(this->name, conf, out); @@ -1123,8 +1339,12 @@ init (xlator_t *this) INIT_LIST_HEAD (&conf->peers); INIT_LIST_HEAD (&conf->volumes); + INIT_LIST_HEAD (&conf->snapshots); + INIT_LIST_HEAD (&conf->missed_snaps_list); + pthread_mutex_init (&conf->mutex, NULL); conf->rpc = rpc; + conf->uds_rpc = uds_rpc; conf->gfs_mgmt = &gd_brick_prog; strncpy (conf->workdir, workdir, PATH_MAX); @@ -1135,6 +1355,8 @@ init (xlator_t *this) glusterd_friend_sm_init (); glusterd_op_sm_init (); glusterd_opinfo_init (); + glusterd_mgmt_v3_lock_init (); + glusterd_txn_opinfo_dict_init (); ret = glusterd_sm_tr_log_init (&conf->op_sm_log, glusterd_op_sm_state_name_get, glusterd_op_sm_event_name_get, @@ -1142,6 +1364,12 @@ init (xlator_t *this) if (ret) goto out; + conf->base_port = GF_IANA_PRIV_PORTS_START; + if (dict_get_uint32(this->options, "base-port", &conf->base_port) == 0) { + gf_log (this->name, GF_LOG_INFO, + "base-port override: %d", conf->base_port); + } + /* Set option to run bricks on valgrind if enabled in glusterd.vol */ conf->valgrind = _gf_false; ret = dict_get_str (this->options, "run-with-valgrind", &valgrind_str); @@ -1241,11 +1469,17 @@ fini (xlator_t *this) goto out; conf = this->private; + + glusterd_stop_uds_listener (this); + FREE (conf->pmap); if (conf->handle) gf_store_handle_destroy (conf->handle); glusterd_sm_tr_log_delete (&conf->op_sm_log); + glusterd_mgmt_v3_lock_fini (); + glusterd_txn_opinfo_dict_fini (); GF_FREE (conf); + this->private = NULL; out: return; @@ -1346,5 +1580,18 @@ struct volume_options options[] = { .description = "Sets the quorum percentage for the trusted " "storage pool." }, + { .key = {"glusterd-sockfile"}, + .type = GF_OPTION_TYPE_PATH, + .description = "The socket file on which glusterd should listen for " + "cli requests. Default is "DEFAULT_GLUSTERD_SOCKFILE "." + }, + { .key = {"base-port"}, + .type = GF_OPTION_TYPE_INT, + .description = "Sets the base port for portmap query" + }, + { .key = {"snap-brick-path"}, + .type = GF_OPTION_TYPE_STR, + .description = "directory where the bricks for the snapshots will be created" + }, { .key = {NULL} }, }; diff --git a/xlators/mgmt/glusterd/src/glusterd.h b/xlators/mgmt/glusterd/src/glusterd.h index 8c043a9a4..3aa395ebc 100644 --- a/xlators/mgmt/glusterd/src/glusterd.h +++ b/xlators/mgmt/glusterd/src/glusterd.h @@ -1,5 +1,5 @@ /* - Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com> + Copyright (c) 2006-2013 Red Hat, Inc. <http://www.redhat.com> This file is part of GlusterFS. This file is licensed to you under your choice of the GNU Lesser @@ -40,7 +40,6 @@ #include "store.h" #define GLUSTERD_MAX_VOLUME_NAME 1000 -#define DEFAULT_LOG_FILE_DIRECTORY DATADIR "/log/glusterfs" #define GLUSTERD_TR_LOG_SIZE 50 #define GLUSTERD_NAME "glusterd" #define GLUSTERD_SOCKET_LISTEN_BACKLOG 128 @@ -48,7 +47,14 @@ #define GLUSTERD_QUORUM_RATIO_KEY "cluster.server-quorum-ratio" #define GLUSTERD_GLOBAL_OPT_VERSION "global-option-version" #define GLUSTERD_COMMON_PEM_PUB_FILE "/geo-replication/common_secret.pem.pub" +#define GEO_CONF_MAX_OPT_VALS 5 +#define GLUSTERD_CREATE_HOOK_SCRIPT "/hooks/1/gsync-create/post/" \ + "S56glusterd-geo-rep-create-post.sh" + +#define GLUSTERD_SNAPS_MAX_HARD_LIMIT 256 +#define GLUSTERD_SNAPS_DEF_SOFT_LIMIT_PERCENT 90 +#define GLUSTERD_SNAPS_MAX_SOFT_LIMIT_PERCENT 100 #define GLUSTERD_SERVER_QUORUM "server" #define FMTSTR_CHECK_VOL_EXISTS "Volume %s does not exist" @@ -69,6 +75,9 @@ struct glusterd_volinfo_; typedef struct glusterd_volinfo_ glusterd_volinfo_t; +struct glusterd_snap_; +typedef struct glusterd_snap_ glusterd_snap_t; + typedef enum glusterd_op_ { GD_OP_NONE = 0, GD_OP_CREATE_VOLUME, @@ -95,10 +104,10 @@ typedef enum glusterd_op_ { GD_OP_LIST_VOLUME, GD_OP_CLEARLOCKS_VOLUME, GD_OP_DEFRAG_BRICK_VOLUME, - GD_OP_BD_OP, GD_OP_COPY_FILE, GD_OP_SYS_EXEC, GD_OP_GSYNC_CREATE, + GD_OP_SNAP, GD_OP_MAX, } glusterd_op_t; @@ -109,8 +118,8 @@ struct glusterd_volgen { }; typedef struct { - struct rpc_clnt *rpc; - gf_boolean_t online; + struct rpc_clnt *rpc; + gf_boolean_t online; } nodesrv_t; typedef struct { @@ -120,37 +129,45 @@ typedef struct { } gd_global_opts_t; typedef struct { - struct _volfile_ctx *volfile; - pthread_mutex_t mutex; - struct list_head peers; - struct list_head xaction_peers; - gf_boolean_t verify_volfile_checksum; - gf_boolean_t trace; - uuid_t uuid; - char workdir[PATH_MAX]; - rpcsvc_t *rpc; - nodesrv_t *shd; - nodesrv_t *nfs; - struct pmap_registry *pmap; - struct list_head volumes; - pthread_mutex_t xprt_lock; - struct list_head xprt_list; - gf_store_handle_t *handle; - gf_timer_t *timer; - glusterd_sm_tr_log_t op_sm_log; - struct rpc_clnt_program *gfs_mgmt; - - struct list_head mount_specs; - gf_boolean_t valgrind; - pthread_t brick_thread; - void *hooks_priv; + struct _volfile_ctx *volfile; + pthread_mutex_t mutex; + struct list_head peers; + struct list_head xaction_peers; + gf_boolean_t verify_volfile_checksum; + gf_boolean_t trace; + uuid_t uuid; + char workdir[PATH_MAX]; + rpcsvc_t *rpc; + nodesrv_t *shd; + nodesrv_t *nfs; + struct pmap_registry *pmap; + struct list_head volumes; + struct list_head snapshots; /*List of snap volumes */ + pthread_mutex_t xprt_lock; + struct list_head xprt_list; + gf_store_handle_t *handle; + gf_timer_t *timer; + glusterd_sm_tr_log_t op_sm_log; + struct rpc_clnt_program *gfs_mgmt; + + struct list_head mount_specs; + gf_boolean_t valgrind; + pthread_t brick_thread; + void *hooks_priv; /* need for proper handshake_t */ - int op_version; /* Starts with 1 for 3.3.0 */ - xlator_t *xl; /* Should be set to 'THIS' before creating thread */ - gf_boolean_t pending_quorum_action; - dict_t *opts; - synclock_t big_lock; - gf_boolean_t restart_done; + int op_version; /* Starts with 1 for 3.3.0 */ + xlator_t *xl; /* Should be set to 'THIS' before creating thread */ + gf_boolean_t pending_quorum_action; + dict_t *opts; + synclock_t big_lock; + gf_boolean_t restart_done; + rpcsvc_t *uds_rpc; /* RPCSVC for the unix domain socket */ + uint32_t base_port; + uint64_t snap_max_hard_limit; + uint64_t snap_max_soft_limit; + char *snap_bricks_directory; + gf_store_handle_t *missed_snaps_list_shandle; + struct list_head missed_snaps_list; } glusterd_conf_t; @@ -160,18 +177,22 @@ typedef enum gf_brick_status { } gf_brick_status_t; struct glusterd_brickinfo { - char hostname[1024]; - char path[PATH_MAX]; - struct list_head brick_list; - uuid_t uuid; - int port; - int rdma_port; - char *logfile; - gf_boolean_t signed_in; + char hostname[1024]; + char path[PATH_MAX]; + char device_path[PATH_MAX]; + struct list_head brick_list; + uuid_t uuid; + int port; + int rdma_port; + char *logfile; + gf_boolean_t signed_in; gf_store_handle_t *shandle; - gf_brick_status_t status; - struct rpc_clnt *rpc; - int decommissioned; + gf_brick_status_t status; + struct rpc_clnt *rpc; + int decommissioned; + char vg[PATH_MAX]; /* FIXME: Use max size for length of vg */ + int caps; /* Capability */ + int32_t snap_status; }; typedef struct glusterd_brickinfo glusterd_brickinfo_t; @@ -194,7 +215,7 @@ struct glusterd_defrag_info_ { int cmd; pthread_t th; gf_defrag_status_t defrag_status; - struct rpc_clnt * rpc; + struct rpc_clnt *rpc; uint32_t connected; char mount[1024]; char databuf[131072]; @@ -227,84 +248,147 @@ struct _auth { typedef struct _auth auth_t; -typedef enum glusterd_vol_backend_ { - GD_VOL_BK_DEFAULT = 0, /* POSIX */ - GD_VOL_BK_BD = 1, -} glusterd_vol_backend_t; +/* Capabilities of xlator */ +#define CAPS_BD 0x00000001 +#define CAPS_THIN 0x00000002 +#define CAPS_OFFLOAD_COPY 0x00000004 +#define CAPS_OFFLOAD_SNAPSHOT 0x00000008 struct glusterd_rebalance_ { - gf_defrag_status_t defrag_status; - uint64_t rebalance_files; - uint64_t rebalance_data; - uint64_t lookedup_files; - uint64_t skipped_files; + gf_defrag_status_t defrag_status; + uint64_t rebalance_files; + uint64_t rebalance_data; + uint64_t lookedup_files; + uint64_t skipped_files; glusterd_defrag_info_t *defrag; - gf_cli_defrag_type defrag_cmd; - uint64_t rebalance_failures; - uuid_t rebalance_id; - double rebalance_time; - glusterd_op_t op; + gf_cli_defrag_type defrag_cmd; + uint64_t rebalance_failures; + uuid_t rebalance_id; + double rebalance_time; + glusterd_op_t op; + dict_t *dict; /* Dict to store misc information + * like list of bricks being removed */ }; typedef struct glusterd_rebalance_ glusterd_rebalance_t; struct glusterd_replace_brick_ { gf_rb_status_t rb_status; - glusterd_brickinfo_t *src_brick; - glusterd_brickinfo_t *dst_brick; + glusterd_brickinfo_t *src_brick; + glusterd_brickinfo_t *dst_brick; uuid_t rb_id; }; typedef struct glusterd_replace_brick_ glusterd_replace_brick_t; struct glusterd_volinfo_ { - char volname[GLUSTERD_MAX_VOLUME_NAME]; - int type; - int brick_count; - struct list_head vol_list; - struct list_head bricks; - glusterd_volume_status status; - int sub_count; /* backward compatibility */ - int stripe_count; - int replica_count; - int subvol_count; /* Number of subvolumes in a + gf_lock_t lock; + char volname[GLUSTERD_MAX_VOLUME_NAME]; + gf_boolean_t is_snap_volume; + glusterd_snap_t *snapshot; + gf_boolean_t is_volume_restored; + char parent_volname[GLUSTERD_MAX_VOLUME_NAME]; + /* In case of a snap volume + i.e (is_snap_volume == TRUE) this + field will contain the name of + the volume which is snapped. In + case of a non-snap volume, this + field will be initialized as N/A */ + int type; + int brick_count; + uint64_t snap_count; + uint64_t snap_max_hard_limit; + struct list_head vol_list; + /* In case of a snap volume + i.e (is_snap_volume == TRUE) this + is linked to glusterd_snap_t->volumes. + In case of a non-snap volume, this is + linked to glusterd_conf_t->volumes */ + struct list_head snapvol_list; + /* This is a current pointer for + glusterd_volinfo_t->snap_volumes */ + struct list_head bricks; + struct list_head snap_volumes; + /* TODO : Need to remove this, as this + * is already part of snapshot object. + */ + glusterd_volume_status status; + int sub_count; /* backward compatibility */ + int stripe_count; + int replica_count; + int subvol_count; /* Number of subvolumes in a distribute volume */ - int dist_leaf_count; /* Number of bricks in one + int dist_leaf_count; /* Number of bricks in one distribute subvolume */ - int port; - gf_store_handle_t *shandle; - gf_store_handle_t *rb_shandle; - gf_store_handle_t *node_state_shandle; + int port; + gf_store_handle_t *shandle; + gf_store_handle_t *rb_shandle; + gf_store_handle_t *node_state_shandle; /* Defrag/rebalance related */ - glusterd_rebalance_t rebal; + glusterd_rebalance_t rebal; /* Replace brick status */ - glusterd_replace_brick_t rep_brick; + glusterd_replace_brick_t rep_brick; - int version; - uint32_t cksum; - gf_transport_type transport_type; - gf_transport_type nfs_transport_type; + int version; + uint32_t cksum; + gf_transport_type transport_type; + gf_transport_type nfs_transport_type; - dict_t *dict; + dict_t *dict; - uuid_t volume_id; - auth_t auth; - char *logdir; + uuid_t volume_id; + auth_t auth; + char *logdir; - dict_t *gsync_slaves; + dict_t *gsync_slaves; - int decommission_in_progress; - xlator_t *xl; + int decommission_in_progress; + xlator_t *xl; - gf_boolean_t memory_accounting; - glusterd_vol_backend_t backend; + gf_boolean_t memory_accounting; + int caps; /* Capability */ - int op_version; - int client_op_version; + int op_version; + int client_op_version; }; +typedef enum gd_snap_status_ { + GD_SNAP_STATUS_NONE, + GD_SNAP_STATUS_INIT, + GD_SNAP_STATUS_IN_USE, + GD_SNAP_STATUS_DECOMMISSION, + GD_SNAP_STATUS_RESTORED, +} gd_snap_status_t; + +struct glusterd_snap_ { + gf_lock_t lock; + struct list_head volumes; + struct list_head snap_list; + char snapname[GLUSTERD_MAX_SNAP_NAME]; + uuid_t snap_id; + char *description; + time_t time_stamp; + gf_boolean_t snap_restored; + gd_snap_status_t snap_status; + gf_store_handle_t *shandle; +}; + +typedef struct glusterd_snap_op_ { + int32_t brick_num; + char *brick_path; + int32_t op; + int32_t status; + struct list_head snap_ops_list; +} glusterd_snap_op_t; + +typedef struct glusterd_missed_snap_ { + char *node_snap_info; + struct list_head missed_snaps; + struct list_head snap_ops; +} glusterd_missed_snap_info; + typedef enum gd_node_type_ { GD_NODE_NONE, GD_NODE_BRICK, @@ -313,6 +397,12 @@ typedef enum gd_node_type_ { GD_NODE_NFS, } gd_node_type; +typedef enum missed_snap_stat { + GD_MISSED_SNAP_NONE, + GD_MISSED_SNAP_PENDING, + GD_MISSED_SNAP_DONE, +} missed_snap_stat; + typedef struct glusterd_pending_node_ { struct list_head list; void *node; @@ -320,6 +410,13 @@ typedef struct glusterd_pending_node_ { int32_t index; } glusterd_pending_node_t; +struct gsync_config_opt_vals_ { + char *op_name; + int no_of_pos_vals; + gf_boolean_t case_sensitive; + char *values[GEO_CONF_MAX_OPT_VALS]; +}; + enum glusterd_op_ret { GLUSTERD_CONNECTION_AWAITED = 100, }; @@ -337,11 +434,18 @@ enum glusterd_vol_comp_status_ { #define GLUSTERD_VOLUME_DIR_PREFIX "vols" #define GLUSTERD_PEER_DIR_PREFIX "peers" #define GLUSTERD_VOLUME_INFO_FILE "info" +#define GLUSTERD_SNAP_INFO_FILE "info" #define GLUSTERD_VOLUME_RBSTATE_FILE "rbstate" #define GLUSTERD_BRICK_INFO_DIR "bricks" #define GLUSTERD_CKSUM_FILE "cksum" #define GLUSTERD_TRASH "trash" #define GLUSTERD_NODE_STATE_FILE "node_state.info" +#define GLUSTERD_MISSED_SNAPS_LIST_FILE "missed_snaps_list" +#define GLUSTERD_VOL_SNAP_DIR_PREFIX "snaps" + +#define GLUSTERD_DEFAULT_SNAPS_BRICK_DIR "/gluster/snaps" +#define GLUSTERD_VAR_RUN_DIR "/var/run" +#define GLUSTERD_RUN_DIR "/run" /* definitions related to replace brick */ #define RB_CLIENT_MOUNTPOINT "rb_mount" @@ -354,14 +458,29 @@ enum glusterd_vol_comp_status_ { typedef ssize_t (*gd_serialize_t) (struct iovec outmsg, void *args); -#define GLUSTERD_GET_VOLUME_DIR(path, volinfo, priv) \ - snprintf (path, PATH_MAX, "%s/vols/%s", priv->workdir,\ - volinfo->volname); +#define GLUSTERD_GET_VOLUME_DIR(path, volinfo, priv) \ + if (volinfo->is_snap_volume) { \ + snprintf (path, PATH_MAX, "%s/snaps/%s/%s", priv->workdir, \ + volinfo->snapshot->snapname, volinfo->volname); \ + } else { \ + snprintf (path, PATH_MAX, "%s/vols/%s", priv->workdir, \ + volinfo->volname); \ + } -#define GLUSTERD_GET_BRICK_DIR(path, volinfo, priv) \ - snprintf (path, PATH_MAX, "%s/%s/%s/%s", priv->workdir,\ - GLUSTERD_VOLUME_DIR_PREFIX, volinfo->volname, \ - GLUSTERD_BRICK_INFO_DIR); +#define GLUSTERD_GET_SNAP_DIR(path, snap, priv) \ + snprintf (path, PATH_MAX, "%s/snaps/%s", priv->workdir, \ + snap->snapname); + +#define GLUSTERD_GET_BRICK_DIR(path, volinfo, priv) \ + if (volinfo->is_snap_volume) { \ + snprintf (path, PATH_MAX, "%s/snaps/%s/%s/%s", priv->workdir, \ + volinfo->snapshot->snapname, volinfo->volname, \ + GLUSTERD_BRICK_INFO_DIR); \ + } else { \ + snprintf (path, PATH_MAX, "%s/%s/%s/%s", priv->workdir, \ + GLUSTERD_VOLUME_DIR_PREFIX, volinfo->volname, \ + GLUSTERD_BRICK_INFO_DIR); \ + } #define GLUSTERD_GET_NFS_DIR(path, priv) \ snprintf (path, PATH_MAX, "%s/nfs", priv->workdir); @@ -381,7 +500,7 @@ typedef ssize_t (*gd_serialize_t) (struct iovec outmsg, void *args); GLUSTERD_GET_VOLUME_DIR (volpath, volinfo, priv); \ GLUSTERD_REMOVE_SLASH_FROM_PATH (brickinfo->path, exp_path); \ snprintf (pidfile, PATH_MAX, "%s/run/%s-%s.pid", \ - volpath, brickinfo->hostname, exp_path); \ + volpath, brickinfo->hostname, exp_path); \ } while (0) #define GLUSTERD_GET_NFS_PIDFILE(pidfile,nfspath) { \ @@ -390,9 +509,9 @@ typedef ssize_t (*gd_serialize_t) (struct iovec outmsg, void *args); } #define GLUSTERD_STACK_DESTROY(frame) do {\ - frame->local = NULL; \ - STACK_DESTROY (frame->root);\ - } while (0) + frame->local = NULL; \ + STACK_DESTROY (frame->root); \ + } while (0) #define GLUSTERD_GET_DEFRAG_DIR(path, volinfo, priv) do { \ char vol_path[PATH_MAX]; \ @@ -414,6 +533,19 @@ typedef ssize_t (*gd_serialize_t) (struct iovec outmsg, void *args); uuid_utoa(MY_UUID)); \ } while (0) +#define GLUSTERD_GET_UUID_NOHYPHEN(ret_string, uuid) do { \ + char *snap_volname_ptr = ret_string; \ + char *snap_volid_ptr = uuid_utoa(uuid); \ + while (*snap_volid_ptr) { \ + if (*snap_volid_ptr == '-') { \ + snap_volid_ptr++; \ + } else { \ + (*snap_volname_ptr++) = \ + (*snap_volid_ptr++); \ + } \ + } \ + *snap_volname_ptr = '\0'; \ + } while (0) int glusterd_uuid_init(); @@ -424,11 +556,11 @@ int glusterd_uuid_generate_save (); static inline unsigned char * __glusterd_uuid() { - glusterd_conf_t *priv = THIS->private; + glusterd_conf_t *priv = THIS->private; - if (uuid_is_null (priv->uuid)) - glusterd_uuid_init(); - return &priv->uuid[0]; + if (uuid_is_null (priv->uuid)) + glusterd_uuid_init(); + return &priv->uuid[0]; } int glusterd_big_locked_notify (struct rpc_clnt *rpc, void *mydata, @@ -476,6 +608,14 @@ int glusterd_op_unlock_send_resp (rpcsvc_request_t *req, int32_t status); int +glusterd_op_mgmt_v3_lock_send_resp (rpcsvc_request_t *req, + uuid_t *txn_id, int32_t status); + +int +glusterd_op_mgmt_v3_unlock_send_resp (rpcsvc_request_t *req, + uuid_t *txn_id, int32_t status); + +int glusterd_op_stage_send_resp (rpcsvc_request_t *req, int32_t op, int32_t status, char *op_errstr, dict_t *rsp_dict); @@ -691,6 +831,9 @@ int glusterd_handle_cli_heal_volume (rpcsvc_request_t *req); int glusterd_handle_cli_list_volume (rpcsvc_request_t *req); +int +glusterd_handle_snapshot (rpcsvc_request_t *req); + /* op-sm functions */ int glusterd_op_stage_heal_volume (dict_t *dict, char **op_errstr); int glusterd_op_heal_volume (dict_t *dict, char **op_errstr); @@ -732,7 +875,6 @@ int glusterd_op_statedump_volume (dict_t *dict, char **op_errstr); int glusterd_op_stage_clearlocks_volume (dict_t *dict, char **op_errstr); int glusterd_op_clearlocks_volume (dict_t *dict, char **op_errstr, dict_t *rsp_dict); -int glusterd_op_stage_bd (dict_t *dict, char **op_errstr); /* misc */ void glusterd_do_replace_brick (void *data); @@ -744,9 +886,67 @@ int glusterd_op_statedump_volume_args_get (dict_t *dict, char **volname, int glusterd_op_gsync_args_get (dict_t *dict, char **op_errstr, char **master, char **slave, char **host_uuid); +int glusterd_stop_volume (glusterd_volinfo_t *volinfo); + /* Synctask part */ int32_t glusterd_op_begin_synctask (rpcsvc_request_t *req, glusterd_op_t op, void *dict); int32_t glusterd_defrag_event_notify_handle (dict_t *dict); + +int32_t +glusterd_txn_opinfo_dict_init (); + +void +glusterd_txn_opinfo_dict_fini (); + +void +glusterd_txn_opinfo_init (); + +/* snapshot */ +glusterd_snap_t* +glusterd_new_snap_object(); + +int32_t +glusterd_list_add_snapvol (glusterd_volinfo_t *origin_vol, + glusterd_volinfo_t *snap_vol); + +glusterd_snap_t* +glusterd_remove_snap_by_id (uuid_t snap_id); + +glusterd_snap_t* +glusterd_remove_snap_by_name (char *snap_name); + +glusterd_snap_t* +glusterd_find_snap_by_name (char *snap_name); + +glusterd_snap_t* +glusterd_find_snap_by_id (uuid_t snap_id); + +int +glusterd_snapshot_prevalidate (dict_t *dict, char **op_errstr, + dict_t *rsp_dict); +int +glusterd_snapshot_brickop (dict_t *dict, char **op_errstr, dict_t *rsp_dict); +int +glusterd_snapshot (dict_t *dict, char **op_errstr, dict_t *rsp_dict); +int +glusterd_snapshot_postvalidate (dict_t *dict, int32_t op_ret, char **op_errstr, + dict_t *rsp_dict); +char * +glusterd_build_snap_device_path (char *device, char *snapname); +int32_t +glusterd_snap_remove (dict_t *rsp_dict, glusterd_snap_t *snap, + gf_boolean_t remove_lvm, gf_boolean_t force); +int32_t +glusterd_snapshot_cleanup (dict_t *dict, char **op_errstr, dict_t *rsp_dict); + +int32_t +glusterd_add_missed_snaps_to_list (dict_t *dict, int32_t missed_snap_count); + +int32_t +glusterd_store_missed_snaps_list (char *missed_info, int32_t brick_num, + char *brick_path, int32_t snap_op, + int32_t snap_status); + #endif |