glusterd : Disallow peer detach if snapshot bricks exist on it

Problem :
- Deploy gluster on 2 nodes, one brick each, one volume replicated
- Create a snapshot
- Lose one server
- Add a replacement peer and new brick with a new IP address
- replace-brick the missing brick onto the new server
  (wait for replication to finish)
- peer detach the old server
- after doing above steps, glusterd fails to restart.

Solution:
  With the fix detach peer will populate an error : "N2 is part of
  existing snapshots. Remove those snapshots before proceeding".
  While doing so we force user to stay with that peer or to delete
  all snapshots.

Change-Id: I3699afb9b2a5f915768b77f885e783bd9b51818c
BUG: 1322145
Signed-off-by: Gaurav Yadav <gyadav@redhat.com>
Reviewed-on: https://review.gluster.org/16907
Smoke: Gluster Build System <jenkins@build.gluster.org>
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>

5 files changed, 94 insertions, 2 deletions

diff --git a/rpc/rpc-lib/src/protocol-common.h b/rpc/rpc-lib/src/protocol-common.h
index 8865baf759b..b5f5bdee50e 100644
--- a/rpc/rpc-lib/src/protocol-common.h
+++ b/rpc/rpc-lib/src/protocol-common.h
@@ -134,6 +134,7 @@ enum gf_deprobe_resp {
         GF_DEPROBE_FRIEND_DOWN,
         GF_DEPROBE_QUORUM_NOT_MET,
         GF_DEPROBE_FRIEND_DETACHING,
+        GF_DEPROBE_SNAP_BRICK_EXIST,
 };
 
 enum gf_cbk_procnum {
diff --git a/tests/bugs/glusterd/bug-1322145-disallow-detatch-peer.t b/tests/bugs/glusterd/bug-1322145-disallow-detatch-peer.t
new file mode 100644
index 00000000000..60eceb4f44d
--- /dev/null
+++ b/tests/bugs/glusterd/bug-1322145-disallow-detatch-peer.t
@@ -0,0 +1,36 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../cluster.rc
+. $(dirname $0)/../../volume.rc
+. $(dirname $0)/../../snapshot.rc
+
+
+cleanup;
+TEST verify_lvm_version
+TEST launch_cluster 3;
+TEST setup_lvm 3;
+
+TEST $CLI_1 peer probe $H2;
+EXPECT_WITHIN $PROBE_TIMEOUT 1 peer_count
+
+TEST $CLI_1 volume create $V0 replica 2 $H1:$L1 $H2:$L2
+EXPECT 'Created' volinfo_field $V0 'Status'
+
+TEST $CLI_1 volume start $V0
+EXPECT 'Started' volinfo_field $V0 'Status'
+
+TEST $CLI_1 snapshot create snap1 $V0 no-timestamp;
+
+kill_glusterd 2
+
+TEST $CLI_1 peer probe $H3;
+EXPECT_WITHIN $PROBE_TIMEOUT 1 peer_count
+
+TEST $CLI_1 volume replace-brick $V0 $H2:$L2 $H3:$L3 commit force
+
+
+# peer hosting snapshotted bricks should not be detachable
+TEST ! $CLI_1 peer detach $H2
+cleanup;
+
diff --git a/xlators/mgmt/glusterd/src/glusterd-handler.c b/xlators/mgmt/glusterd/src/glusterd-handler.c
index a630d55d9a3..8368354e65f 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handler.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handler.c
@@ -1318,6 +1318,8 @@ __glusterd_handle_cli_deprobe (rpcsvc_request_t *req)
         int                           flags = 0;
         glusterd_volinfo_t         *volinfo = NULL;
         glusterd_volinfo_t             *tmp = NULL;
+        glusterd_snap_t           *snapinfo = NULL;
+        glusterd_snap_t            *tmpsnap = NULL;
 
         this = THIS;
         GF_ASSERT (this);
@@ -1402,14 +1404,21 @@ __glusterd_handle_cli_deprobe (rpcsvc_request_t *req)
         */
         cds_list_for_each_entry_safe (volinfo, tmp, &priv->volumes,
                                       vol_list) {
-                ret = glusterd_friend_contains_vol_bricks (volinfo,
-                                                           uuid);
+                ret = glusterd_friend_contains_vol_bricks (volinfo, uuid);
                 if (ret == 1) {
                         op_errno = GF_DEPROBE_BRICK_EXIST;
                         goto out;
                 }
         }
 
+        cds_list_for_each_entry_safe (snapinfo, tmpsnap, &priv->snapshots,
+                                      snap_list) {
+                ret = glusterd_friend_contains_snap_bricks (snapinfo, uuid);
+                if (ret == 1) {
+                        op_errno = GF_DEPROBE_SNAP_BRICK_EXIST;
+                        goto out;
+                }
+        }
         if (!(flags & GF_CLI_FLAG_OP_FORCE)) {
                 if (glusterd_is_any_volume_in_server_quorum (this) &&
                     !does_gd_meet_server_quorum (this)) {
@@ -3954,6 +3963,12 @@ set_deprobe_error_str (int op_ret, int op_errno, char *op_errstr, char *errstr,
                                           "%s exist in cluster", hostname);
                                 break;
 
+                        case GF_DEPROBE_SNAP_BRICK_EXIST:
+                                snprintf (errstr, len, "%s is part of existing "
+                                          "snapshot. Remove those snapshots "
+                                          "before proceeding ", hostname);
+                                break;
+
                         case GF_DEPROBE_FRIEND_DOWN:
                                 snprintf (errstr, len, "One of the peers is "
                                           "probably down. Check with "
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index fbec4fcafb8..89bfa3d7358 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -7777,6 +7777,42 @@ glusterd_friend_contains_vol_bricks (glusterd_volinfo_t *volinfo,
         return ret;
 }
 
+/* Checks if the given peer contains bricks belonging to the given volume.
+ * Returns,
+ *   2 - if peer contains all the bricks
+ *   1 - if peer contains at least 1 brick
+ *   0 - if peer contains no bricks
+ */
+int
+glusterd_friend_contains_snap_bricks (glusterd_snap_t *snapinfo,
+                                     uuid_t friend_uuid)
+{
+        int                                ret = -1;
+        glusterd_volinfo_t            *volinfo = NULL;
+        glusterd_brickinfo_t        *brickinfo = NULL;
+        int                              count = 0;
+
+        GF_VALIDATE_OR_GOTO ("glusterd", snapinfo, out);
+
+        cds_list_for_each_entry (volinfo, &snapinfo->volumes, vol_list) {
+                cds_list_for_each_entry (brickinfo, &volinfo->bricks,
+                                                          brick_list) {
+                        if (!gf_uuid_compare (brickinfo->uuid, friend_uuid)) {
+                                count++;
+                        }
+                }
+        }
+
+        if (count > 0)
+                ret = 1;
+        else
+                ret = 0;
+
+out:
+        gf_msg_debug (THIS->name, 0, "Returning %d", ret);
+        return ret;
+}
+
 /* Cleanup the stale volumes left behind in the cluster. The volumes which are
  * contained completely within the detached peer are stale with respect to the
  * cluster.
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h
index db13c4c8ad4..1a0376da5cd 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.h
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.h
@@ -411,6 +411,10 @@ glusterd_is_brick_decommissioned (glusterd_volinfo_t *volinfo, char *hostname,
 int
 glusterd_friend_contains_vol_bricks (glusterd_volinfo_t *volinfo,
                                      uuid_t friend_uuid);
+
+int
+glusterd_friend_contains_snap_bricks (glusterd_snap_t *snapinfo,
+                                     uuid_t friend_uuid);
 int
 glusterd_friend_remove_cleanup_vols (uuid_t uuid);