diff options
author | Kaushal M <kaushal@redhat.com> | 2013-08-28 17:16:03 +0530 |
---|---|---|
committer | Vijay Bellur <vbellur@redhat.com> | 2013-09-10 09:09:18 -0700 |
commit | 8565d383a11d52b2db649f0c277d6f9ef1e87623 (patch) | |
tree | 8624c4ecfaf2f36814bd7e34f688c934db9a7bb4 | |
parent | 9d4ea7a870668d9af9bbd415ff9ad3aace59e170 (diff) |
glusterd: Allow bumping down a peer's op-version during probev3.4.1qa1
Backport of http://review.gluster.org/5715
Earlier, a peer running a higher op-version couldn't be probed into a
cluster running at a lower op-version. This created issues when trying
to expand an upgraded cluster. This patch changes this behaviour.
The cluster no longer rejects a peer being probed if its op-version is
higher than the cluster op-version. The peer will reduce its op-version
if it doesn't have any volumes. If the peer contains volumes and needs
to reduce its op-version, it fails the handshake and the probe fails.
BUG: 1005038
Change-Id: Iabe790a9f826a4ac63d379eeeba01efcfef01f4d
Signed-off-by: Kaushal M <kaushal@redhat.com>
Reviewed-on: http://review.gluster.org/5834
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>
-rw-r--r-- | doc/release-notes/3.4.0.md | 13 | ||||
-rw-r--r-- | xlators/mgmt/glusterd/src/glusterd-handshake.c | 24 |
2 files changed, 9 insertions, 28 deletions
diff --git a/doc/release-notes/3.4.0.md b/doc/release-notes/3.4.0.md index 15aff92fa..b906e6a94 100644 --- a/doc/release-notes/3.4.0.md +++ b/doc/release-notes/3.4.0.md @@ -165,16 +165,3 @@ More details can be found [here] - For Block Device translator based volumes open-behind translator at the client side needs to be disabled. - -- More information on this can be found in this mailing list - [post](http://www.gluster.org/pipermail/gluster-users/2013-July/036750.html) - A freshly installed peer cannot be probed into a freshly upgraded cluster. - This is a limitation brought about by the introduction of op-versions. - There are two ways to workaround this issue, - * enable any new feature (for eg. open-behind) on a volume in the cluster, - and then perform the probe. - Or, if you don't want to enable any new features - * delete the 'operating-version' line from /var/lib/glusterd/glusterd.info - on the newly installed peer and restart glusterd on the peer. - If the file doesn't exist (because of lazy uuid generation), try to probe - the peer from the cluster (which will fail) to get the file created. diff --git a/xlators/mgmt/glusterd/src/glusterd-handshake.c b/xlators/mgmt/glusterd/src/glusterd-handshake.c index 9124c46ee..faf05b4e2 100644 --- a/xlators/mgmt/glusterd/src/glusterd-handshake.c +++ b/xlators/mgmt/glusterd/src/glusterd-handshake.c @@ -408,12 +408,16 @@ gd_validate_cluster_op_version (xlator_t *this, int cluster_op_version, goto out; } - if (cluster_op_version < conf->op_version) { + /* The peer can only reduce its op-version when it doesn't have any + * volumes. Reducing op-version when it already contains volumes can + * lead to inconsistencies in the cluster + */ + if ((cluster_op_version < conf->op_version) && + !list_empty (&conf->volumes)) { gf_log (this->name, GF_LOG_ERROR, - "operating version %d is less than the currently " - "running version (%d) on the machine (as per peer " - "request from %s)", cluster_op_version, - conf->op_version, peerid); + "cannot reduce operating version to %d from current " + "version %d as volumes exist (as per peer request from " + "%s)", cluster_op_version, conf->op_version, peerid); goto out; } @@ -734,16 +738,6 @@ gd_validate_peer_op_version (xlator_t *this, glusterd_peerinfo_t *peerinfo, goto out; } - /* If peer is already operating at a higher op_version reject it. - * Cluster cannot be moved to higher op_version to accomodate a peer. - */ - if (peer_op_version > conf->op_version) { - ret = gf_asprintf (errstr, "Peer %s is already at a higher " - "op-version", peerinfo->hostname); - ret = -1; - goto out; - } - ret = 0; out: gf_log (this->name , GF_LOG_DEBUG, "Peer %s %s", peerinfo->hostname, |