summaryrefslogtreecommitdiffstats
path: root/xlators/mgmt
diff options
context:
space:
mode:
authorKrishnan Parthasarathi <kp@gluster.com>2012-04-03 15:37:25 +0530
committerVijay Bellur <vijay@gluster.com>2012-04-12 23:07:08 -0700
commit8a45a0e480f7e8c6ea1195f77ce3810d4817dc37 (patch)
tree9b50579b53228f6a605b4c74e36fe0fb7a3827e2 /xlators/mgmt
parentb337b755325f75a6fcf65616eaf4467b70b8b245 (diff)
glusterd: Added volume-id to 'op' dictionary
Volume-id passed in op dictionary would help detect possible split brains among peers in a cluster. The idea is to check if the volume's id and the vol-id that was passed are equal. ie, same volume name, but different volume id indicate that glusterd 'metadata' of one of the participating peers is stale or there is a split brain. This is over and above the existing checksum based validation of peer supplied cluster 'metadata' (ie, volume info file). Change-Id: I1049ef249e417e540ccb4243e450f92fcd0f46f9 BUG: 797734 Signed-off-by: Krishnan Parthasarathi <kp@gluster.com> Reviewed-on: http://review.gluster.com/3083 Tested-by: Gluster Build System <jenkins@build.gluster.com> Reviewed-by: Jeff Darcy <jdarcy@redhat.com> Reviewed-by: Vijay Bellur <vijay@gluster.com>
Diffstat (limited to 'xlators/mgmt')
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-brick-ops.c8
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-op-sm.c53
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-utils.c28
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-utils.h2
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-volume-ops.c24
5 files changed, 115 insertions, 0 deletions
diff --git a/xlators/mgmt/glusterd/src/glusterd-brick-ops.c b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c
index aa57341ffd7..300c40a8d5e 100644
--- a/xlators/mgmt/glusterd/src/glusterd-brick-ops.c
+++ b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c
@@ -1065,6 +1065,10 @@ glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr)
goto out;
}
+ ret = glusterd_validate_volume_id (dict, volinfo);
+ if (ret)
+ goto out;
+
if (glusterd_is_rb_ongoing (volinfo)) {
snprintf (msg, sizeof (msg), "Replace brick is in progress on "
"volume %s. Please retry after replace-brick "
@@ -1201,6 +1205,10 @@ glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr)
goto out;
}
+ ret = glusterd_validate_volume_id (dict, volinfo);
+ if (ret)
+ goto out;
+
if (glusterd_is_rb_ongoing (volinfo)) {
snprintf (msg, sizeof (msg), "Replace brick is in progress on "
"volume %s. Please retry after replace-brick "
diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
index a577e161c17..1825065a4b4 100644
--- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c
+++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
@@ -504,6 +504,7 @@ glusterd_op_stage_reset_volume (dict_t *dict, char **op_errstr)
char msg[2048] = {0};
char *key = NULL;
char *key_fixed = NULL;
+ glusterd_volinfo_t *volinfo = NULL;
ret = dict_get_str (dict, "volname", &volname);
@@ -522,6 +523,13 @@ glusterd_op_stage_reset_volume (dict_t *dict, char **op_errstr)
ret = -1;
goto out;
}
+ ret = glusterd_volinfo_find (volname, &volinfo);
+ if (ret)
+ goto out;
+
+ ret = glusterd_validate_volume_id (dict, volinfo);
+ if (ret)
+ goto out;
ret = dict_get_str (dict, "key", &key);
if (ret) {
@@ -568,6 +576,7 @@ glusterd_op_stage_sync_volume (dict_t *dict, char **op_errstr)
gf_boolean_t exists = _gf_false;
glusterd_peerinfo_t *peerinfo = NULL;
char msg[2048] = {0,};
+ glusterd_volinfo_t *volinfo = NULL;
ret = dict_get_str (dict, "hostname", &hostname);
if (ret) {
@@ -607,6 +616,13 @@ glusterd_op_stage_sync_volume (dict_t *dict, char **op_errstr)
ret = -1;
goto out;
}
+ ret = glusterd_volinfo_find (volname, &volinfo);
+ if (ret)
+ goto out;
+
+ ret = glusterd_validate_volume_id (dict, volinfo);
+ if (ret)
+ goto out;
} else {
ret = 0;
}
@@ -664,6 +680,10 @@ glusterd_op_stage_status_volume (dict_t *dict, char **op_errstr)
goto out;
}
+ ret = glusterd_validate_volume_id (dict, volinfo);
+ if (ret)
+ goto out;
+
ret = glusterd_is_volume_started (volinfo);
if (!ret) {
snprintf (msg, sizeof (msg), "Volume %s is not started",
@@ -799,6 +819,10 @@ glusterd_op_stage_stats_volume (dict_t *dict, char **op_errstr)
goto out;
}
+ ret = glusterd_validate_volume_id (dict, volinfo);
+ if (ret)
+ goto out;
+
ret = dict_get_int32 (dict, "op", &stats_op);
if (ret) {
snprintf (msg, sizeof (msg), "Volume profile op get failed");
@@ -1771,6 +1795,9 @@ glusterd_op_build_payload (dict_t **req)
void *ctx = NULL;
dict_t *req_dict = NULL;
glusterd_op_t op = GD_OP_NONE;
+ char *volname = NULL;
+ char *volid = NULL;
+ glusterd_volinfo_t *volinfo = NULL;
GF_ASSERT (req);
@@ -1820,6 +1847,32 @@ glusterd_op_build_payload (dict_t **req)
case GD_OP_DEFRAG_BRICK_VOLUME:
{
dict_t *dict = ctx;
+ ret = dict_get_str (dict, "volname", &volname);
+ if (ret) {
+ gf_log (THIS->name, GF_LOG_CRITICAL,
+ "volname is not present in "
+ "operation ctx");
+ goto out;
+ }
+ ret = glusterd_volinfo_find (volname, &volinfo);
+ if (ret) {
+ gf_log (THIS->name, GF_LOG_ERROR,
+ "volume %s not present in "
+ "the cluster", volname);
+ goto out;
+ }
+ volid = gf_strdup (uuid_utoa (volinfo->volume_id));
+ if (!volid) {
+ ret = -1;
+ goto out;
+ }
+ ret = dict_set_dynstr (dict, "vol-id", volid);
+ if (ret) {
+ gf_log (THIS->name, GF_LOG_ERROR,
+ "Failed to set volume id in "
+ "dictionary");
+ goto out;
+ }
dict_copy (dict, req_dict);
}
break;
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index 817cbbc323e..7198a130775 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -5328,3 +5328,31 @@ glusterd_is_local_brick (xlator_t *this, glusterd_volinfo_t *volinfo,
out:
return local;
}
+int
+glusterd_validate_volume_id (dict_t *op_dict, glusterd_volinfo_t *volinfo)
+{
+ int ret = -1;
+ char *volid_str = NULL;
+ uuid_t vol_uid = {0, };
+
+ ret = dict_get_str (op_dict, "vol-id", &volid_str);
+ if (ret) {
+ gf_log (THIS->name, GF_LOG_ERROR, "Failed to get volume id");
+ goto out;
+ }
+ ret = uuid_parse (volid_str, vol_uid);
+ if (ret) {
+ gf_log (THIS->name, GF_LOG_ERROR, "Failed to parse uuid");
+ goto out;
+ }
+
+ if (uuid_compare (vol_uid, volinfo->volume_id)) {
+ gf_log (THIS->name, GF_LOG_ERROR, "Volume ids are different. "
+ "Possibly a split brain among peers.");
+ ret = -1;
+ goto out;
+ }
+
+out:
+ return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h
index 5491c7a0068..602dc780fef 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.h
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.h
@@ -432,4 +432,6 @@ glusterd_get_brickinfo_by_position (glusterd_volinfo_t *volinfo, uint32_t pos);
gf_boolean_t
glusterd_is_local_brick (xlator_t *this, glusterd_volinfo_t *volinfo,
glusterd_brickinfo_t *brickinfo);
+int
+glusterd_validate_volume_id (dict_t *op_dict, glusterd_volinfo_t *volinfo);
#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
index b30c4631806..85e7abba853 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
@@ -830,6 +830,10 @@ glusterd_op_stage_start_volume (dict_t *dict, char **op_errstr)
if (ret)
goto out;
+ ret = glusterd_validate_volume_id (dict, volinfo);
+ if (ret)
+ goto out;
+
list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
ret = glusterd_resolve_brick (brickinfo);
if (ret) {
@@ -905,6 +909,10 @@ glusterd_op_stage_stop_volume (dict_t *dict, char **op_errstr)
if (ret)
goto out;
+ ret = glusterd_validate_volume_id (dict, volinfo);
+ if (ret)
+ goto out;
+
/* If 'force' flag is given, no check is required */
if (flags & GF_CLI_FLAG_OP_FORCE)
goto out;
@@ -1003,6 +1011,10 @@ glusterd_op_stage_delete_volume (dict_t *dict, char **op_errstr)
if (ret)
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 has been started."
"Volume needs to be stopped before deletion.",
@@ -1056,6 +1068,10 @@ glusterd_op_stage_heal_volume (dict_t *dict, char **op_errstr)
goto out;
}
+ ret = glusterd_validate_volume_id (dict, volinfo);
+ if (ret)
+ goto out;
+
if (!glusterd_is_volume_replicate (volinfo)) {
ret = -1;
snprintf (msg, sizeof (msg), "Volume %s is not of type "
@@ -1142,6 +1158,10 @@ glusterd_op_stage_statedump_volume (dict_t *dict, char **op_errstr)
goto out;
}
+ ret = glusterd_validate_volume_id (dict, volinfo);
+ if (ret)
+ goto out;
+
is_running = glusterd_is_volume_started (volinfo);
if (!is_running) {
snprintf (msg, sizeof(msg), "Volume %s is not in a started"
@@ -1209,6 +1229,10 @@ glusterd_op_stage_clearlocks_volume (dict_t *dict, char **op_errstr)
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);