Tier: Stop tierd for detach start

Problem: tierd was stopped only after detach commit
This makes the detach take a longer time. The detach
demotes the files to the cold brick and if the promotion
frequency is hit, then the tierd starts to promote files to
hot tier again.

Fix: stop tierd after detach start so the files get
demoted faster.

Note: the is_tier_enabled was not maintained properly.
That has been fixed too. some code clean up has been done.

Signed-off-by: hari gowtham <hgowtham@redhat.com>

Change-Id: I532f7410cea04fbb960105483810ea3560ca149b
BUG: 1446381

7 files changed, 158 insertions, 173 deletions

diff --git a/xlators/mgmt/glusterd/src/glusterd-messages.h b/xlators/mgmt/glusterd/src/glusterd-messages.h
index cf84cd30f11..a112d72d822 100644
--- a/xlators/mgmt/glusterd/src/glusterd-messages.h
+++ b/xlators/mgmt/glusterd/src/glusterd-messages.h
@@ -41,7 +41,7 @@
 
 #define GLUSTERD_COMP_BASE      GLFS_MSGID_GLUSTERD
 
-#define GLFS_NUM_MESSAGES       613
+#define GLFS_NUM_MESSAGES       614
 
 #define GLFS_MSGID_END          (GLUSTERD_COMP_BASE + GLFS_NUM_MESSAGES + 1)
 /* Messaged with message IDs */
@@ -4953,6 +4953,14 @@
  */
 #define GD_MSG_CHANGELOG_GET_FAIL (GLUSTERD_COMP_BASE + 613)
 
+/*!
+ * @messageid
+ * @diagnosis
+ * @recommendedaction
+ *
+ */
+#define GD_MSG_MANAGER_FUNCTION_FAILED  (GLUSTERD_COMP_BASE + 614)
+
 /*------------*/
 
 #define glfs_msg_end_x GLFS_MSGID_END, "Invalid: End of messages"
diff --git a/xlators/mgmt/glusterd/src/glusterd-mgmt.c b/xlators/mgmt/glusterd/src/glusterd-mgmt.c
index 3fb90766082..a1596f0944d 100644
--- a/xlators/mgmt/glusterd/src/glusterd-mgmt.c
+++ b/xlators/mgmt/glusterd/src/glusterd-mgmt.c
@@ -539,8 +539,6 @@ gd_mgmt_v3_post_validate_fn (glusterd_op_t op, int32_t op_ret, dict_t *dict,
                                 goto out;
                         }
 
-                        volinfo->is_tier_enabled = _gf_true;
-
                         if (ret) {
                                 gf_msg (this->name, GF_LOG_ERROR, errno,
                                         GD_MSG_DICT_SET_FAILED, "dict set "
diff --git a/xlators/mgmt/glusterd/src/glusterd-tier.c b/xlators/mgmt/glusterd/src/glusterd-tier.c
index f9a1d35803d..5df87fdfa33 100644
--- a/xlators/mgmt/glusterd/src/glusterd-tier.c
+++ b/xlators/mgmt/glusterd/src/glusterd-tier.c
@@ -244,116 +244,6 @@ glusterd_handle_tier (rpcsvc_request_t *req)
         return glusterd_big_locked_handler (req, __glusterd_handle_tier);
 }
 
-
-static int
-glusterd_manage_tier (glusterd_volinfo_t *volinfo, int opcode)
-{
-        int              ret   = -1;
-        xlator_t         *this = NULL;
-        glusterd_conf_t  *priv = NULL;
-
-        this = THIS;
-        GF_VALIDATE_OR_GOTO (THIS->name, this, out);
-        GF_VALIDATE_OR_GOTO (this->name, volinfo, out);
-        priv = this->private;
-        GF_VALIDATE_OR_GOTO (this->name, priv, out);
-
-        switch (opcode) {
-        case GF_DEFRAG_CMD_START_TIER:
-        case GF_DEFRAG_CMD_STOP_TIER:
-                ret = volinfo->tierd.svc.manager (&(volinfo->tierd.svc),
-                                                volinfo, PROC_START_NO_WAIT);
-                break;
-        default:
-                ret = 0;
-                break;
-        }
-
-out:
-        return ret;
-
-}
-
-static int
-glusterd_tier_enable (glusterd_volinfo_t *volinfo, char **op_errstr)
-{
-        int32_t                 ret                     = -1;
-        xlator_t                *this                   = NULL;
-        int32_t                 tier_online             = -1;
-        char                    pidfile[PATH_MAX]       = {0};
-        int32_t                 pid                     = -1;
-        glusterd_conf_t         *priv                   = NULL;
-
-        this = THIS;
-
-        GF_VALIDATE_OR_GOTO (THIS->name, this, out);
-        GF_VALIDATE_OR_GOTO (this->name, volinfo, out);
-        GF_VALIDATE_OR_GOTO (this->name, op_errstr, out);
-        priv = this->private;
-        GF_VALIDATE_OR_GOTO (this->name, priv, out);
-
-        if (glusterd_is_volume_started (volinfo) == 0) {
-                *op_errstr = gf_strdup ("Volume is stopped, start volume "
-                                        "to enable tier.");
-                ret = -1;
-                goto out;
-        }
-
-        GLUSTERD_GET_TIER_PID_FILE(pidfile, volinfo, priv);
-        tier_online = gf_is_service_running (pidfile, &pid);
-
-        if (tier_online) {
-                *op_errstr = gf_strdup ("tier is already enabled");
-                ret = -1;
-                goto out;
-        }
-
-        volinfo->is_tier_enabled = _gf_true;
-
-        ret = 0;
-out:
-        if (ret && op_errstr && !*op_errstr)
-                gf_asprintf (op_errstr, "Enabling tier on volume %s has been "
-                             "unsuccessful", volinfo->volname);
-        return ret;
-}
-
-static int
-glusterd_tier_disable (glusterd_volinfo_t *volinfo, char **op_errstr)
-{
-        int32_t                 ret                     = -1;
-        xlator_t                *this                   = NULL;
-        int32_t                 tier_online             = -1;
-        char                    pidfile[PATH_MAX]       = {0};
-        int32_t                 pid                     = -1;
-        glusterd_conf_t         *priv                   = NULL;
-
-        this = THIS;
-
-        GF_VALIDATE_OR_GOTO (THIS->name, this, out);
-        GF_VALIDATE_OR_GOTO (this->name, volinfo, out);
-        GF_VALIDATE_OR_GOTO (this->name, op_errstr, out);
-        priv = this->private;
-
-        GLUSTERD_GET_TIER_PID_FILE(pidfile, volinfo, priv);
-        tier_online = gf_is_service_running (pidfile, &pid);
-
-        if (!tier_online) {
-                *op_errstr = gf_strdup ("tier is already disabled");
-                ret = -1;
-                goto out;
-        }
-
-        volinfo->is_tier_enabled = _gf_false;
-
-        ret = 0;
-out:
-        if (ret && op_errstr && !*op_errstr)
-                gf_asprintf (op_errstr, "Disabling tier volume %s has "
-                             "been unsuccessful", volinfo->volname);
-        return ret;
-}
-
 int
 glusterd_op_remove_tier_brick (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
 {
@@ -455,6 +345,19 @@ glusterd_op_remove_tier_brick (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
                                         continue;
                                 brickinfo->decommissioned = 0;
                         }
+                        volinfo->tier.op = GD_OP_DETACH_NOT_STARTED;
+                        ret = volinfo->tierd.svc.manager (&(volinfo->tierd.svc),
+                                                          volinfo,
+                                                          PROC_START_NO_WAIT);
+                        if (ret) {
+                                gf_msg (this->name, GF_LOG_ERROR, 0,
+                                        GD_MSG_MANAGER_FUNCTION_FAILED,
+                                        "Calling manager for tier "
+                                        "failed on volume: %s for "
+                                        "detach stop", volinfo->volname);
+                                goto out;
+                        }
+
                         ret = glusterd_create_volfiles_and_notify_services
                                 (volinfo);
 
@@ -473,22 +376,24 @@ glusterd_op_remove_tier_brick (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
                                         "failed to store volinfo");
                                 goto out;
                         }
-                        ret = glusterd_tierdsvc_restart ();
-                        if (ret) {
-                                gf_msg (this->name, GF_LOG_ERROR, 0,
-                                        GD_MSG_TIERD_START_FAIL,
-                                        "Couldn't restart tierd for "
-                                        "vol: %s", volinfo->volname);
-                                goto out;
-                        }
-
-                        volinfo->tier.op = GD_OP_DETACH_NOT_STARTED;
                         ret = 0;
                         goto out;
 
 
 
         case GF_DEFRAG_CMD_DETACH_START:
+                        volinfo->tier.op = GD_OP_DETACH_TIER;
+                        svc = &(volinfo->tierd.svc);
+                        ret = svc->manager (svc, volinfo,
+                                        PROC_START_NO_WAIT);
+                        if (ret) {
+                                gf_msg (this->name, GF_LOG_ERROR, 0,
+                                        GD_MSG_MANAGER_FUNCTION_FAILED,
+                                        "calling manager for tier "
+                                        "failed on volume: %s for "
+                                        "detach start", volname);
+                                goto out;
+                        }
                         ret = dict_get_str (dict, GF_REMOVE_BRICK_TID_KEY,
                                             &task_id_str);
                         if (ret) {
@@ -510,8 +415,6 @@ glusterd_op_remove_tier_brick (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
                         }
                         force = 0;
 
-                        volinfo->tier.op = GD_OP_DETACH_TIER;
-                        volinfo->tier.defrag_status = GF_DEFRAG_STATUS_STARTED;
                         break;
 
         case GF_DEFRAG_CMD_DETACH_COMMIT:
@@ -529,6 +432,19 @@ glusterd_op_remove_tier_brick (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
                         /* Fall through */
 
         case GF_DEFRAG_CMD_DETACH_COMMIT_FORCE:
+                        if (cmd == GF_DEFRAG_CMD_DETACH_COMMIT_FORCE) {
+                                svc = &(volinfo->tierd.svc);
+                                ret = svc->manager (svc, volinfo,
+                                                PROC_START_NO_WAIT);
+                                if (ret) {
+                                        gf_msg (this->name, GF_LOG_ERROR, 0,
+                                                GD_MSG_MANAGER_FUNCTION_FAILED,
+                                                "calling manager for tier "
+                                                "failed on volume: %s for "
+                                                "commit force", volname);
+                                        goto out;
+                                }
+                        }
                         glusterd_op_perform_detach_tier (volinfo);
                         detach_commit = 1;
 
@@ -707,11 +623,6 @@ glusterd_op_remove_tier_brick (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
         if (cmd == GF_DEFRAG_CMD_DETACH_START &&
                         volinfo->status == GLUSTERD_STATUS_STARTED) {
 
-                svc = &(volinfo->tierd.svc);
-                ret = svc->reconfigure (volinfo);
-                if (ret)
-                        goto out;
-
                 ret = glusterd_svcs_reconfigure ();
                 if (ret) {
                         gf_msg (this->name, GF_LOG_WARNING, 0,
@@ -780,6 +691,7 @@ glusterd_op_tier_start_stop (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
         glusterd_conf_t         *priv                   = NULL;
         int32_t                 pid                     = -1;
         char                    pidfile[PATH_MAX]       = {0};
+        int                     is_force                = 0;
 
         this = THIS;
         GF_VALIDATE_OR_GOTO (THIS->name, this, out);
@@ -821,24 +733,48 @@ glusterd_op_tier_start_stop (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
         if (!retval)
                 goto out;
 
+        if (glusterd_is_volume_started (volinfo) == 0) {
+                *op_errstr = gf_strdup ("Volume is stopped, start "
+                                "volume to enable/disable tier.");
+                ret = -1;
+                goto out;
+        }
+
+        GLUSTERD_GET_TIER_PID_FILE(pidfile, volinfo, priv);
+
         switch (cmd) {
         case GF_DEFRAG_CMD_START_TIER:
-                GLUSTERD_GET_TIER_PID_FILE(pidfile, volinfo, priv);
                 /* we check if its running and skip so that we dont get a
                  * failure during force start
                  */
-                if (gf_is_service_running (pidfile, &pid))
-                        goto out;
-                ret = glusterd_tier_enable (volinfo, op_errstr);
-                if (ret < 0)
-                        goto out;
-                glusterd_store_perform_node_state_store (volinfo);
+                ret = dict_get_int32 (dict, "force", &is_force);
+                if (ret) {
+                        gf_msg_debug (this->name, 0, "Unable to get is_force"
+                                        " from dict");
+                }
+                ret = dict_set_int32 (volinfo->dict, "force", is_force);
+                if (ret) {
+                        gf_msg_debug (this->name, errno, "Unable to set"
+                                        " is_force to dict");
+                }
+
+                if (!is_force) {
+                        if (gf_is_service_running (pidfile, &pid)) {
+                                gf_asprintf (op_errstr, "Tier is already "
+                                             "enabled on volume %s." ,
+                                             volinfo->volname);
+                                goto out;
+                        }
+                }
+
                 break;
 
         case GF_DEFRAG_CMD_STOP_TIER:
-                ret = glusterd_tier_disable (volinfo, op_errstr);
-                if (ret < 0)
+                if (!gf_is_service_running (pidfile, &pid)) {
+                        gf_asprintf (op_errstr, "Tier is alreaady disabled on "
+                                     "volume %s.", volinfo->volname);
                         goto out;
+                }
                 break;
         default:
                 gf_asprintf (op_errstr, "tier command failed. Invalid "
@@ -847,7 +783,8 @@ glusterd_op_tier_start_stop (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
                 goto out;
         }
 
-        ret = glusterd_manage_tier (volinfo, cmd);
+        ret = volinfo->tierd.svc.manager (&(volinfo->tierd.svc),
+                                          volinfo, PROC_START_NO_WAIT);
         if (ret)
                 goto out;
 
@@ -984,6 +921,19 @@ glusterd_op_stage_tier (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
                                 "start validate failed");
                         goto out;
                 }
+                if (volinfo->tier.op == GD_OP_DETACH_TIER) {
+                        snprintf (msg, sizeof (msg), "A detach tier task "
+                                  "exists for volume %s. Either commit it"
+                                  " or stop it before starting a new task.",
+                                  volinfo->volname);
+                        gf_msg (this->name, GF_LOG_ERROR, 0,
+                                GD_MSG_OLD_REMOVE_BRICK_EXISTS,
+                                "Earlier detach-tier"
+                                " task exists for volume %s.",
+                                volinfo->volname);
+                        ret = -1;
+                        goto out;
+                }
                 break;
 
         case GF_DEFRAG_CMD_STOP_TIER:
diff --git a/xlators/mgmt/glusterd/src/glusterd-tierd-svc.c b/xlators/mgmt/glusterd/src/glusterd-tierd-svc.c
index bb2cbff6e93..378ecdb7a0d 100644
--- a/xlators/mgmt/glusterd/src/glusterd-tierd-svc.c
+++ b/xlators/mgmt/glusterd/src/glusterd-tierd-svc.c
@@ -72,6 +72,8 @@ glusterd_tierdsvc_init (void *data)
         notify = glusterd_svc_common_rpc_notify;
         glusterd_store_perform_node_state_store (volinfo);
 
+        volinfo->type = GF_CLUSTER_TYPE_TIER;
+
         glusterd_svc_build_tierd_rundir (volinfo, rundir, sizeof (rundir));
         glusterd_svc_create_rundir (rundir);
 
@@ -150,6 +152,7 @@ glusterd_tierdsvc_manager (glusterd_svc_t *svc, void *data, int flags)
         int                 ret     = 0;
         xlator_t           *this    = THIS;
         glusterd_volinfo_t *volinfo = NULL;
+        int                 is_force = 0;
 
         volinfo = data;
         GF_VALIDATE_OR_GOTO (this->name, data, out);
@@ -169,25 +172,29 @@ glusterd_tierdsvc_manager (glusterd_svc_t *svc, void *data, int flags)
                 }
         }
 
-        ret = glusterd_is_tierd_enabled (volinfo);
-        if (ret == -1) {
-                gf_msg (this->name, GF_LOG_ERROR, 0,
-                        GD_MSG_VOLINFO_GET_FAIL, "Failed to read volume "
-                        "options");
-                goto out;
+        ret = dict_get_int32 (volinfo->dict, "force", &is_force);
+        if (ret) {
+                gf_msg_debug (this->name, errno, "Unable to get"
+                              " is_force from dict");
         }
 
+        if (is_force)
+                ret = 1;
+        else
+                ret = (glusterd_is_tierd_supposed_to_be_enabled (volinfo));
+
         if (ret) {
                 if (!glusterd_is_volume_started (volinfo)) {
                         if (glusterd_proc_is_running (&svc->proc)) {
                                 ret = svc->stop (svc, SIGTERM);
                                 if (ret)
                                         gf_msg (this->name, GF_LOG_ERROR, 0,
-                                                GD_MSG_TIERD_STOP_FAIL,
+                                                GD_MSG_SNAPD_STOP_FAIL,
                                                 "Couldn't stop tierd for "
                                                 "volume: %s",
                                                 volinfo->volname);
                         } else {
+                                /* Since tierd is not running set ret to 0 */
                                 ret = 0;
                         }
                         goto out;
@@ -209,6 +216,7 @@ glusterd_tierdsvc_manager (glusterd_svc_t *svc, void *data, int flags)
                                 "tierd for volume: %s", volinfo->volname);
                         goto out;
                 }
+                volinfo->is_tier_enabled = _gf_true;
 
                 glusterd_volinfo_ref (volinfo);
                 ret = glusterd_conn_connect (&(svc->conn));
@@ -216,16 +224,19 @@ glusterd_tierdsvc_manager (glusterd_svc_t *svc, void *data, int flags)
                         glusterd_volinfo_unref (volinfo);
                         goto out;
                 }
-
-        } else if (glusterd_proc_is_running (&svc->proc)) {
-                ret = svc->stop (svc, SIGTERM);
-                if (ret) {
-                        gf_msg (this->name, GF_LOG_ERROR, 0,
-                                GD_MSG_TIERD_STOP_FAIL,
-                                "Couldn't stop tierd for volume: %s",
-                                volinfo->volname);
-                        goto out;
+        } else {
+                if (glusterd_proc_is_running (&svc->proc)) {
+                        ret = svc->stop (svc, SIGTERM);
+                        if (ret) {
+                                gf_msg (this->name, GF_LOG_ERROR, 0,
+                                        GD_MSG_TIERD_STOP_FAIL,
+                                        "Couldn't stop tierd for volume: %s",
+                                        volinfo->volname);
+                                goto out;
+                        }
+                        volinfo->is_tier_enabled = _gf_false;
                 }
+                ret = 0;
         }
 
 out:
@@ -362,7 +373,6 @@ out:
         return ret;
 }
 
-
 int
 glusterd_tierdsvc_restart ()
 {
@@ -380,15 +390,18 @@ glusterd_tierdsvc_restart ()
         cds_list_for_each_entry (volinfo, &conf->volumes, vol_list) {
                 /* Start per volume tierd svc */
                 if (volinfo->status == GLUSTERD_STATUS_STARTED &&
-                    glusterd_is_tierd_enabled (volinfo)) {
+                    volinfo->type == GF_CLUSTER_TYPE_TIER) {
                         svc = &(volinfo->tierd.svc);
-                        ret = svc->manager (svc, volinfo, PROC_START_NO_WAIT);
-                        if (ret) {
-                                gf_msg (this->name, GF_LOG_ERROR, 0,
-                                        GD_MSG_TIERD_START_FAIL,
-                                        "Couldn't restart tierd for "
-                                        "vol: %s", volinfo->volname);
-                                goto out;
+                           if (volinfo->tier.op != GD_OP_DETACH_TIER) {
+                                ret = svc->manager (svc, volinfo,
+                                                PROC_START_NO_WAIT);
+                                if (ret) {
+                                        gf_msg (this->name, GF_LOG_ERROR, 0,
+                                                GD_MSG_TIERD_START_FAIL,
+                                                "Couldn't restart tierd for "
+                                                "vol: %s", volinfo->volname);
+                                        goto out;
+                                }
                         }
                 }
         }
@@ -419,7 +432,7 @@ glusterd_tierdsvc_reconfigure (void *data)
         this = THIS;
         GF_VALIDATE_OR_GOTO (THIS->name, this, out);
 
-        if (glusterd_is_tierd_enabled (volinfo))
+        if (!glusterd_is_tierd_enabled (volinfo))
                 goto manager;
         /*
          * Check both OLD and NEW volfiles, if they are SAME by size
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index fe96d6be094..066d2f72f81 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -12174,6 +12174,16 @@ glusterd_is_volume_inode_quota_enabled (glusterd_volinfo_t *volinfo)
 }
 
 int
+glusterd_is_tierd_supposed_to_be_enabled (glusterd_volinfo_t *volinfo)
+{
+        if ((volinfo->type != GF_CLUSTER_TYPE_TIER) ||
+                        (volinfo->tier.op == GD_OP_DETACH_TIER))
+                return _gf_false;
+        else
+                return _gf_true;
+}
+
+int
 glusterd_is_tierd_enabled (glusterd_volinfo_t *volinfo)
 {
         return volinfo->is_tier_enabled;
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h
index e76ee63edf1..6111ea1100f 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.h
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.h
@@ -628,6 +628,9 @@ int
 glusterd_is_tierd_enabled (glusterd_volinfo_t *volinfo);
 
 int
+glusterd_is_tierd_supposed_to_be_enabled (glusterd_volinfo_t *volinfo);
+
+int
 glusterd_is_volume_quota_enabled (glusterd_volinfo_t *volinfo);
 
 int
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
index 0db91a30fff..a87dfc39eb7 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
@@ -2729,6 +2729,16 @@ glusterd_stop_volume (glusterd_volinfo_t *volinfo)
                 }
         }
 
+        /* call tier manager before the voluem status is set as stopped
+         * as tier uses that as a check in the manager
+         * */
+        if (volinfo->type == GF_CLUSTER_TYPE_TIER) {
+                svc = &(volinfo->tierd.svc);
+                ret = svc->manager (svc, volinfo, PROC_START_NO_WAIT);
+                if (ret)
+                        goto out;
+        }
+
         glusterd_set_volume_status (volinfo, GLUSTERD_STATUS_STOPPED);
 
         ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT);
@@ -2746,13 +2756,6 @@ glusterd_stop_volume (glusterd_volinfo_t *volinfo)
                         goto out;
         }
 
-        if (volinfo->type == GF_CLUSTER_TYPE_TIER) {
-                svc = &(volinfo->tierd.svc);
-                ret = svc->manager (svc, volinfo, PROC_START_NO_WAIT);
-                if (ret)
-                        goto out;
-        }
-
         ret = glusterd_svcs_manager (volinfo);
         if (ret) {
                 gf_msg (this->name, GF_LOG_ERROR, 0,