Description of problem: We have seen below issue while testing open/lock states recovery after IP failover in nfs-ganesha HA setup - The cluster has not been put into grace before IP fails over. There seem to be a timing issue there. "pcs constraint colocation order" seem to have not completely solved it.
yes, there's a race between setting grace and failing over the virt IP. See http://review.gluster.org/10490
The required changes to fix this bug have not made it into glusterfs-3.7.1. This bug is now getting tracked for glusterfs-3.7.2.
Not sure why I created 1227028 on release-3.7 branch. I may have meant for it to be downstream. 1219485 = main 1227028 = release-3.7 1224618 = downstream *** This bug has been marked as a duplicate of bug 1227028 ***