glusterd: spawn daemons from init() on a single or two node setup

Allow glusterd to spawn the daemons at the time of initialization when peer
count is less than 2. This is required if user wants to set up a two node
cluster with out server side quorum and want the bricks to come up on a node
where the other node is down, however the behaviour will be overriden when
server side quorum is enabled.

Change-Id: I21118e996655822467eaf329f638eb9a8bf8b7d5
BUG: 1352277
Signed-off-by: Atin Mukherjee <amukherj@redhat.com>
Reviewed-on: http://review.gluster.org/14848
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>

4 files changed, 67 insertions, 6 deletions

diff --git a/tests/bugs/glusterd/bug-1352277-spawn-daemons-on-two-node-setup.t b/tests/bugs/glusterd/bug-1352277-spawn-daemons-on-two-node-setup.t
new file mode 100644
index 00000000000..53d8d34160e
--- /dev/null
+++ b/tests/bugs/glusterd/bug-1352277-spawn-daemons-on-two-node-setup.t
@@ -0,0 +1,37 @@
+#!/bin/bash
+
+# Test case for checking whether the brick process(es) come up on a two node
+# cluster if one of them is already down and other is going through a restart
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+. $(dirname $0)/../../cluster.rc
+
+cleanup;
+
+TEST launch_cluster 2
+
+TEST $CLI_1 peer probe $H2;
+EXPECT_WITHIN $PROBE_TIMEOUT 1  peer_count
+
+# Lets create the volume
+TEST $CLI_1 volume create $V0 $H1:$B1/${V0}1 $H2:$B2/${V0}2
+
+# Start the volume
+TEST $CLI_1 volume start $V0
+
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status_1 $V0 $H1 $B1/${V0}1
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status_1 $V0 $H2 $B2/${V0}2
+
+# Bring down all the gluster processes
+TEST killall_gluster
+
+#Bring back 1st glusterd and check whether the brick process comes back
+TEST $glusterd_1
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "1" brick_up_status_1 $V0 $H1 $B1/${V0}1
+
+#Enabling quorum should bring down the brick
+TEST $CLI_1 volume set $V0 cluster.server-quorum-type server
+EXPECT_WITHIN $PROCESS_UP_TIMEOUT "0" brick_up_status_1 $V0 $H1 $B1/${V0}1
+
+cleanup;
diff --git a/xlators/mgmt/glusterd/src/glusterd-peer-utils.c b/xlators/mgmt/glusterd/src/glusterd-peer-utils.c
index 4131296ef12..1a97111d0f5 100644
--- a/xlators/mgmt/glusterd/src/glusterd-peer-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-peer-utils.c
@@ -1034,3 +1034,25 @@ glusterd_peerinfo_find_by_generation (uint32_t generation) {
                         generation);
         return found;
 }
+
+int
+glusterd_get_peers_count () {
+        int                  count = 0;
+        xlator_t            *this  = NULL;
+        glusterd_conf_t     *conf  = NULL;
+        glusterd_peerinfo_t *peer  = NULL;
+
+        this = THIS;
+        GF_VALIDATE_OR_GOTO ("glusterd", this, out);
+
+        conf = this->private;
+        GF_VALIDATE_OR_GOTO (this->name, conf, out);
+
+        rcu_read_lock ();
+        cds_list_for_each_entry_rcu (peer, &conf->peers, uuid_list)
+                count++;
+        rcu_read_unlock ();
+
+out:
+        return count;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-peer-utils.h b/xlators/mgmt/glusterd/src/glusterd-peer-utils.h
index 9332cf2ea02..e74d1ed9536 100644
--- a/xlators/mgmt/glusterd/src/glusterd-peer-utils.h
+++ b/xlators/mgmt/glusterd/src/glusterd-peer-utils.h
@@ -87,4 +87,7 @@ gd_add_peer_detail_to_dict (glusterd_peerinfo_t *peerinfo, dict_t *friends,
                             int count);
 glusterd_peerinfo_t *
 glusterd_peerinfo_find_by_generation (uint32_t generation);
+
+int
+glusterd_get_peers_count ();
 #endif /* _GLUSTERD_PEER_UTILS_H */
diff --git a/xlators/mgmt/glusterd/src/glusterd.c b/xlators/mgmt/glusterd/src/glusterd.c
index d35ca8dda09..d5002dfe1f1 100644
--- a/xlators/mgmt/glusterd/src/glusterd.c
+++ b/xlators/mgmt/glusterd/src/glusterd.c
@@ -1743,13 +1743,12 @@ init (xlator_t *this)
         if (ret < 0)
                 goto out;
 
-        /* If there are no 'friends', this would be the best time to
-         * spawn process/bricks that may need (re)starting since last
-         * time (this) glusterd was up.*/
-
-        if (cds_list_empty (&conf->peers)) {
+        /* If the peer count is less than 2 then this would be the best time to
+         * spawn process/bricks that may need (re)starting since last time
+         * (this) glusterd was up. */
+        if (glusterd_get_peers_count () < 2)
                 glusterd_launch_synctask (glusterd_spawn_daemons, NULL);
-        }
+
         ret = glusterd_options_init (this);
         if (ret < 0)
                 goto out;