ctdb-recoverd: Call an election when the recovery lock is lost
authorMartin Schwenke <martin@meltin.net>
Thu, 8 Nov 2018 04:49:30 +0000 (15:49 +1100)
committerAmitay Isaacs <amitay@gmail.com>
Tue, 18 Dec 2018 02:36:47 +0000 (13:36 +1100)
commit82bc27f62eb44396523e4511e268a3008e9b13b7
treeee91796a32ff183071b8083f4a7824723baf8f6e
parentc5fd87def80b619fbb8b28a795b8ce0e36c0d732
ctdb-recoverd: Call an election when the recovery lock is lost

The lock may have been lost due to a failure in the underlying locking
mechanism.  This could be due to quorum loss or similar.  It is best
to call an election to confirm that this node should still be master.
At worst, the node will reelect itself, fail to take the lock and then
ban itself.  This is a suitable outcome for a node that has been
partitioned from others in the cluster.

Signed-off-by: Martin Schwenke <martin@meltin.net>
Reviewed-by: Amitay Isaacs <amitay@gmail.com>
ctdb/server/ctdb_recoverd.c