As of now there is no support for glusterfs NLM failover as the new server to which it fails over does not know which clients were previously accessing the failed server to send NSM notifications to. We need to figure out how we can fix this behavior.
This is not reasonably solvable within Gluster/NFS. However, the move to NFS-Ganesha with HA through pacemaker/corosync includes this feature. http://www.gluster.org/community/documentation/index.php/Features/HA_for_ganesha *** This bug has been marked as a duplicate of bug 1188184 ***