Nodes receive "reservation conflict" errors when cluster daemons are stopped manually or when the fence_scsi device is removed from the configuration in a RHEL 5, 6, or 7 High Availability cluster

Issue

• I get "reservation conflict" errors when accessing shared storage devices after stopping the cluster daemons on all nodes.
• The cluster doesn't appear to be removing SCSI reservations from our disks when we stop the cluster
• After stopping the cluster services cman, clvmd, rgmanager, on a cluster node, the SCSI reservation keys for that node are still present on SAN devices
• We've removed the fence_scsi device from our cluster configuration because we're switching to a different method, but we still get "reservation conflict" errors on the nodes even after restarting the cluster