glusterd: Don't start bricks if server quorum is not met

Upon glusterd restart if it is observered that the server quorum
isn't met anymore due to changes to the "server-quorum-ratio"
global option, the bricks should be stopped if they are running.
Also if glusterd has been restarted, and if server quorum is not
applicable for a volume, do not restart the bricks corresponding
to the volume to make sure that bricks that have been brought
down purposely, say for maintenance, are not brought up. This
commit moves this check that was previously inside
"glusterd_spawn_daemons" to "glusterd_restart_bricks" instead.

> Change-Id: I0a44a2e7cad0739ed7d56d2d67ab58058716de6b
> BUG: 1345727
> Signed-off-by: Samikshan Bairagya <samikshan@gmail.com>
> Reviewed-on: http://review.gluster.org/14758
> Smoke: Gluster Build System <jenkins@build.gluster.org>
> NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
> CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
> Reviewed-by: Jeff Darcy <jdarcy@redhat.com>

(cherry picked from commit 807b9a135d697f175fc9933f1d23fb67b0cc6c7d)

Change-Id: I0a44a2e7cad0739ed7d56d2d67ab58058716de6b
BUG: 1353814
Signed-off-by: Samikshan Bairagya <samikshan@gmail.com>
Reviewed-on: http://review.gluster.org/14876
Smoke: Gluster Build System <jenkins@build.gluster.org>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>

1 files changed, 59 insertions, 11 deletions

diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index d85c8947802..02f6185d482 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -3050,14 +3050,10 @@ int
 glusterd_spawn_daemons (void *opaque)
 {
         glusterd_conf_t *conf = THIS->private;
-        gf_boolean_t    start_bricks = !conf->restart_done;
         int             ret             = -1;
 
         synclock_lock (&conf->big_lock);
-        if (start_bricks) {
-                glusterd_restart_bricks (conf);
-                conf->restart_done = _gf_true;
-        }
+        glusterd_restart_bricks (conf);
         glusterd_restart_gsyncds (conf);
         glusterd_restart_rebalance (conf);
         ret = glusterd_snapdsvc_restart ();
@@ -4214,6 +4210,24 @@ out:
 }
 
 int
+glusterd_get_global_server_quorum_ratio (dict_t *opts, double *quorum)
+{
+        int      ret        = -1;
+        char    *quorum_str = NULL;
+
+        ret = dict_get_str (opts, GLUSTERD_QUORUM_RATIO_KEY, &quorum_str);
+        if (ret)
+                goto out;
+
+        ret = gf_string2percent (quorum_str, quorum);
+        if (ret)
+                goto out;
+        ret = 0;
+out:
+        return ret;
+}
+
+int
 glusterd_get_global_opt_version (dict_t *opts, uint32_t *version)
 {
         int     ret = -1;
@@ -4260,6 +4274,8 @@ glusterd_import_global_opts (dict_t *friend_data)
         int             count = 0;
         uint32_t        local_version = 0;
         uint32_t        remote_version = 0;
+        double          old_quorum = 0.0;
+        double          new_quorum = 0.0;
 
         this = THIS;
         conf = this->private;
@@ -4283,19 +4299,41 @@ glusterd_import_global_opts (dict_t *friend_data)
                 goto out;
         }
 
+        /* Not handling ret since server-quorum-ratio might not yet be set */
+        ret = glusterd_get_global_server_quorum_ratio (conf->opts,
+                                                       &old_quorum);
+        ret = glusterd_get_global_server_quorum_ratio (import_options,
+                                                       &new_quorum);
+
         ret = glusterd_get_global_opt_version (conf->opts, &local_version);
         if (ret)
                 goto out;
         ret = glusterd_get_global_opt_version (import_options, &remote_version);
         if (ret)
                 goto out;
+
         if (remote_version > local_version) {
                 ret = glusterd_store_options (this, import_options);
                 if (ret)
                         goto out;
                 dict_unref (conf->opts);
                 conf->opts = dict_ref (import_options);
+
+                /* If server quorum ratio has changed, restart bricks to
+                 * recompute if quorum is met. If quorum is not met bricks are
+                 * not started and those already running are stopped
+                 */
+                if (old_quorum != new_quorum) {
+                        ret = glusterd_restart_bricks (conf);
+                        if (ret) {
+                                gf_msg ("glusterd", GF_LOG_INFO, 0,
+                                        GD_MSG_SERVER_QUORUM_NOT_MET,
+                                        "Restarting bricks failed");
+                                goto out;
+                        }
+                }
         }
+
         ret = 0;
 out:
         if (import_options)
@@ -4321,7 +4359,6 @@ glusterd_compare_friend_data (dict_t *peer_data, int32_t *status,
 
         priv = this->private;
         GF_ASSERT (priv);
-
         ret = glusterd_import_global_opts (peer_data);
         if (ret) {
                 gf_msg (this->name, GF_LOG_ERROR, 0,
@@ -4883,7 +4920,7 @@ glusterd_restart_bricks (glusterd_conf_t *conf)
                         volinfo->volname);
 
                 /* Check the quorum, if quorum is not met, don't start the
-                   bricks
+                   bricks. Stop bricks in case they are running.
                 */
                 ret = check_quorum_for_brick_start (volinfo, node_quorum);
                 if (ret == 0) {
@@ -4891,11 +4928,22 @@ glusterd_restart_bricks (glusterd_conf_t *conf)
                                 GD_MSG_SERVER_QUORUM_NOT_MET, "Skipping brick "
                                 "restart for volume %s as quorum is not met",
                                 volinfo->volname);
+                        (void) glusterd_stop_bricks (volinfo);
                         continue;
-                }
-                cds_list_for_each_entry (brickinfo, &volinfo->bricks,
-                                         brick_list) {
-                        glusterd_brick_start (volinfo, brickinfo, _gf_false);
+                } else if (ret == 2 && conf->restart_done == _gf_true) {
+                        /* If glusterd has been restarted and quorum is not
+                         * applicable then do not restart the bricks as this
+                         * might start bricks brought down purposely, say for
+                         * maintenance
+                         */
+                        continue;
+                } else {
+                        cds_list_for_each_entry (brickinfo, &volinfo->bricks,
+                                                 brick_list) {
+                                glusterd_brick_start (volinfo, brickinfo,
+                                                     _gf_false);
+                        }
+                        conf->restart_done = _gf_true;
                 }
         }