glusterd: add counter support for tiered volumes

        back port of http://review.gluster.org/10292

This fix adds support to view the number of promoted or demoted
files from the cli. The mechanism is isolmorphic to checking
the status of volumes being rebalanced.

gluster volume rebalance <vol> tier status

 >Change-Id: I1b11ca27355ceec36c488967c23531202030e205
 >BUG: 1213063
 >Signed-off-by: Mohammed Rafi KC <rkavunga@redhat.com>
 >Signed-off-by: Dan Lambright <dlambrig@redhat.com>
 >Reviewed-on: http://review.gluster.org/10292
 >Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
 >Tested-by: Gluster Build System <jenkins@build.gluster.com>

Change-Id: I543e886f17132b544274c83fdecca5a8da9d092a
BUG: 1221477
Signed-off-by: Mohammed Rafi KC <rkavunga@redhat.com>
Signed-off-by: Dan Lambright <dlambrig@redhat.com>
Reviewed-on: http://review.gluster.org/10775
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>

6 files changed, 100 insertions, 79 deletions

diff --git a/cli/src/cli-rpc-ops.c b/cli/src/cli-rpc-ops.c
index a6315e28e43..f3ed8adfdc2 100644
--- a/cli/src/cli-rpc-ops.c
+++ b/cli/src/cli-rpc-ops.c
@@ -1602,15 +1602,11 @@ gf_cli_print_tier_status (dict_t *dict, enum gf_task_types task_type)
         int                ret          = -1;
         int                count        = 0;
         int                i            = 1;
+        uint64_t           promoted     = 0;
+        uint64_t           demoted      = 0;
         char               key[256]     = {0,};
-        gf_defrag_status_t status_rcd   = GF_DEFRAG_STATUS_NOT_STARTED;
-        uint64_t           files        = 0;
-        uint64_t           size         = 0;
-        uint64_t           lookup       = 0;
         char               *node_name   = NULL;
-        uint64_t           failures     = 0;
-        uint64_t           skipped      = 0;
-        double             elapsed      = 0;
+        gf_defrag_status_t status_rcd   = GF_DEFRAG_STATUS_NOT_STARTED;
         char               *status_str  = NULL;
         char               *size_str    = NULL;
 
@@ -1620,105 +1616,52 @@ gf_cli_print_tier_status (dict_t *dict, enum gf_task_types task_type)
                 goto out;
         }
 
+        cli_out ("%-20s %-20s %-20s %-20s", "Node", "Promoted files",
+                 "Demoted files", "Status");
+        cli_out ("%-20s %-20s %-20s %-20s", "---------", "---------",
+                 "---------", "---------");
 
-        cli_out ("%40s %16s %13s %13s %13s %13s %20s %18s", "Node",
-                 "Rebalanced-files", "size", "scanned", "failures", "skipped",
-                 "status", "run time in secs");
-        cli_out ("%40s %16s %13s %13s %13s %13s %20s %18s", "---------",
-                 "-----------", "-----------", "-----------", "-----------",
-                 "-----------", "------------", "--------------");
         for (i = 1; i <= count; i++) {
                 /* Reset the variables to prevent carryover of values */
                 node_name = NULL;
-                files = 0;
-                size = 0;
-                lookup = 0;
-                skipped = 0;
-                status_str = NULL;
-                elapsed = 0;
+                promoted = 0;
+                demoted = 0;
 
-                /* Check if status is NOT_STARTED, and continue early */
                 memset (key, 0, 256);
-                snprintf (key, 256, "status-%d", i);
-                ret = dict_get_int32 (dict, key, (int32_t *)&status_rcd);
-                if (ret) {
-                        gf_log ("cli", GF_LOG_TRACE, "failed to get status");
-                        goto out;
-                }
-                if (GF_DEFRAG_STATUS_NOT_STARTED == status_rcd)
-                        continue;
-
-
                 snprintf (key, 256, "node-name-%d", i);
                 ret = dict_get_str (dict, key, &node_name);
                 if (ret)
                         gf_log ("cli", GF_LOG_TRACE, "failed to get node-name");
 
                 memset (key, 0, 256);
-                snprintf (key, 256, "files-%d", i);
-                ret = dict_get_uint64 (dict, key, &files);
-                if (ret)
-                        gf_log ("cli", GF_LOG_TRACE,
-                                "failed to get file count");
-
-                memset (key, 0, 256);
-                snprintf (key, 256, "size-%d", i);
-                ret = dict_get_uint64 (dict, key, &size);
-                if (ret)
-                        gf_log ("cli", GF_LOG_TRACE,
-                                "failed to get size of xfer");
-
-                memset (key, 0, 256);
-                snprintf (key, 256, "lookups-%d", i);
-                ret = dict_get_uint64 (dict, key, &lookup);
+                snprintf (key, 256, "promoted-%d", i);
+                ret = dict_get_uint64 (dict, key, &promoted);
                 if (ret)
                         gf_log ("cli", GF_LOG_TRACE,
-                                "failed to get lookedup file count");
+                                "failed to get promoted count");
 
                 memset (key, 0, 256);
-                snprintf (key, 256, "failures-%d", i);
-                ret = dict_get_uint64 (dict, key, &failures);
+                snprintf (key, 256, "demoted-%d", i);
+                ret = dict_get_uint64 (dict, key, &demoted);
                 if (ret)
                         gf_log ("cli", GF_LOG_TRACE,
-                                "failed to get failures count");
+                                "failed to get demoted count");
 
                 memset (key, 0, 256);
-                snprintf (key, 256, "skipped-%d", i);
-                ret = dict_get_uint64 (dict, key, &skipped);
+                snprintf (key, 256, "status-%d", i);
+                ret = dict_get_int32 (dict, key, (int32_t *)&status_rcd);
                 if (ret)
                         gf_log ("cli", GF_LOG_TRACE,
-                                "failed to get skipped count");
-
-                /* For remove-brick include skipped count into failure count*/
-                if (task_type != GF_TASK_TYPE_REBALANCE) {
-                        failures += skipped;
-                        skipped = 0;
-                }
-
-                memset (key, 0, 256);
-                snprintf (key, 256, "run-time-%d", i);
-                ret = dict_get_double (dict, key, &elapsed);
-                if (ret)
-                        gf_log ("cli", GF_LOG_TRACE, "failed to get run-time");
+                                "failed to get status");
 
                 /* Check for array bound */
                 if (status_rcd >= GF_DEFRAG_STATUS_MAX)
                         status_rcd = GF_DEFRAG_STATUS_MAX;
 
                 status_str = cli_vol_task_status_str[status_rcd];
-                size_str = gf_uint64_2human_readable(size);
-                if (size_str) {
-                        cli_out ("%40s %16"PRIu64 " %13s" " %13"PRIu64 " %13"
-                                 PRIu64" %13"PRIu64 " %20s %18.2f", node_name,
-                                 files, size_str, lookup, failures, skipped,
-                                 status_str, elapsed);
-                } else {
-                        cli_out ("%40s %16"PRIu64 " %13"PRIu64 " %13"PRIu64
-                                 " %13"PRIu64" %13"PRIu64 " %20s %18.2f",
-                                 node_name, files, size, lookup, failures,
-                                 skipped, status_str, elapsed);
-                }
-                GF_FREE(size_str);
+                cli_out ("%-20s %-20"PRIu64" %-20"PRIu64" %-20s",
+                         node_name, promoted, demoted, status_str);
+
         }
 out:
         return ret;
diff --git a/xlators/cluster/dht/src/dht-common.c b/xlators/cluster/dht/src/dht-common.c
index 20ad4f2bd72..8309b317017 100644
--- a/xlators/cluster/dht/src/dht-common.c
+++ b/xlators/cluster/dht/src/dht-common.c
@@ -6482,7 +6482,8 @@ dht_notify (xlator_t *this, int event, void *data, ...)
                 {
                         if (defrag->is_exiting)
                                 goto unlock;
-                        if (cmd == GF_DEFRAG_CMD_STATUS)
+                        if ((cmd == GF_DEFRAG_CMD_STATUS) ||
+                            (cmd == GF_DEFRAG_CMD_STATUS_TIER))
                                 gf_defrag_status_get (defrag, output);
                         else if (cmd == GF_DEFRAG_CMD_START_DETACH_TIER)
                                 gf_defrag_start_detach_tier(defrag);
diff --git a/xlators/cluster/dht/src/dht-rebalance.c b/xlators/cluster/dht/src/dht-rebalance.c
index 9c42af934e4..fcb005ecc3d 100644
--- a/xlators/cluster/dht/src/dht-rebalance.c
+++ b/xlators/cluster/dht/src/dht-rebalance.c
@@ -2911,6 +2911,8 @@ gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict)
         uint64_t lookup = 0;
         uint64_t failures = 0;
         uint64_t skipped = 0;
+        uint64_t promoted = 0;
+        uint64_t demoted = 0;
         char     *status = "";
         double   elapsed = 0;
         struct timeval end = {0,};
@@ -2928,6 +2930,8 @@ gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict)
         lookup = defrag->num_files_lookedup;
         failures = defrag->total_failures;
         skipped = defrag->skipped;
+        promoted = defrag->total_files_promoted;
+        demoted = defrag->total_files_demoted;
 
         gettimeofday (&end, NULL);
 
@@ -2936,6 +2940,16 @@ gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict)
         if (!dict)
                 goto log;
 
+        ret = dict_set_uint64 (dict, "promoted", promoted);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set promoted count");
+
+        ret = dict_set_uint64 (dict, "demoted", demoted);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set demoted count");
+
         ret = dict_set_uint64 (dict, "files", files);
         if (ret)
                 gf_log (THIS->name, GF_LOG_WARNING,
diff --git a/xlators/cluster/dht/src/tier.c b/xlators/cluster/dht/src/tier.c
index 2526604eaba..bf481601bd8 100644
--- a/xlators/cluster/dht/src/tier.c
+++ b/xlators/cluster/dht/src/tier.c
@@ -365,6 +365,12 @@ tier_migrate_using_query_file (void *_args)
                                 per_link_status = -1;
                                 goto error;
                         }
+
+                        if (query_cbk_args->is_promotion)
+                                defrag->total_files_promoted++;
+                        else
+                                defrag->total_files_demoted++;
+
 error:
 
                         loc_wipe(&loc);
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index 04e5707058a..9445de77127 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -6707,6 +6707,8 @@ glusterd_defrag_volume_status_update (glusterd_volinfo_t *volinfo,
         uint64_t                        skipped = 0;
         xlator_t                       *this = NULL;
         double                          run_time = 0;
+        uint64_t                        promoted = 0;
+        uint64_t                        demoted = 0;
 
         this = THIS;
 
@@ -6740,6 +6742,16 @@ glusterd_defrag_volume_status_update (glusterd_volinfo_t *volinfo,
                 gf_log (this->name, GF_LOG_TRACE,
                         "failed to get skipped count");
 
+        ret = dict_get_uint64 (rsp_dict, "promoted", &promoted);
+        if (ret)
+                gf_log (this->name, GF_LOG_TRACE,
+                        "failed to get promoted count");
+
+        ret = dict_get_uint64 (rsp_dict, "demoted", &demoted);
+        if (ret)
+                gf_log (this->name, GF_LOG_TRACE,
+                        "failed to get demoted count");
+
         ret = dict_get_double (rsp_dict, "run-time", &run_time);
         if (ret)
                 gf_log (this->name, GF_LOG_TRACE,
@@ -6759,6 +6771,10 @@ glusterd_defrag_volume_status_update (glusterd_volinfo_t *volinfo,
                 volinfo->rebal.skipped_files = skipped;
         if (run_time)
                 volinfo->rebal.rebalance_time = run_time;
+        if (promoted)
+                volinfo->tier_info.promoted = promoted;
+        if (demoted)
+                volinfo->tier_info.demoted = demoted;
 
         return ret;
 }
@@ -7863,6 +7879,31 @@ glusterd_volume_rebalance_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict)
                 }
         }
 
+        memset (key, 0, 256);
+        snprintf (key, 256, "demoted-%d", index);
+        ret = dict_get_uint64 (rsp_dict, key, &value);
+        if (!ret) {
+                memset (key, 0, 256);
+                snprintf (key, 256, "demoted-%d", current_index);
+                ret = dict_set_uint64 (ctx_dict, key, value);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_DEBUG,
+                                "failed to set demoted count");
+                }
+        }
+        memset (key, 0, 256);
+        snprintf (key, 256, "promoted-%d", index);
+        ret = dict_get_uint64 (rsp_dict, key, &value);
+        if (!ret) {
+                memset (key, 0, 256);
+                snprintf (key, 256, "promoted-%d", current_index);
+                ret = dict_set_uint64 (ctx_dict, key, value);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_DEBUG,
+                                "failed to set promoted count");
+                }
+        }
+
         ret = 0;
 
 out:
@@ -8467,6 +8508,20 @@ glusterd_defrag_volume_node_rsp (dict_t *req_dict, dict_t *rsp_dict,
                 gf_log (THIS->name, GF_LOG_ERROR,
                         "failed to set run-time");
 
+        memset (key, 0 , 256);
+        snprintf (key, 256, "promoted-%d", i);
+        ret = dict_set_uint64 (op_ctx, key, volinfo->tier_info.promoted);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "failed to set lookedup file count");
+
+        memset (key, 0 , 256);
+        snprintf (key, 256, "demoted-%d", i);
+        ret = dict_set_uint64 (op_ctx, key, volinfo->tier_info.demoted);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "failed to set lookedup file count");
+
 out:
         return ret;
 }
diff --git a/xlators/mgmt/glusterd/src/glusterd.h b/xlators/mgmt/glusterd/src/glusterd.h
index 5341192e84a..384b6f4186e 100644
--- a/xlators/mgmt/glusterd/src/glusterd.h
+++ b/xlators/mgmt/glusterd/src/glusterd.h
@@ -314,6 +314,8 @@ typedef struct tier_info_ {
         int                       hot_type;
         int                       hot_brick_count;
         int                       hot_replica_count;
+        int                       promoted;
+        int                       demoted;
 } gd_tier_info_t;
 
 struct glusterd_volinfo_ {