stonith fencing of a node using fence_scsi fails with the logs showing "no reservation exists on device <device>" in a RHEL 6 or 7 High Availability cluster with pacemaker

Issue

• Fencing with fence_scsi is failing with "no reservation exists" errors in the logs

Aug 23 10:03:23 [5168] node1 stonith-ng:     info: call_remote_stonith:  Requesting that node1 perform op reboot node2 with FENCE_dev_scsi for crmd.14471 (72s)
Aug 23 10:03:38 [5168] node1 stonith-ng:    error: log_operation:        Operation 'reboot' [5009] (call 4 from crmd.14471) for host 'node2' with device 'FENCE_dev_scsi' returned: -201 (Generic Pacemaker error)
Aug 23 10:03:38 [5168] node1 stonith-ng:  warning: log_operation:        FENCE_dev_scsi:5009 [ Aug 23 10:03:36 fence_scsi: [debug] no reservation exists on device /dev/dm-196 ]

• My cluster locks up after a node reboots or stops communicating, and I see fence_scsi errors in the log regarding no reservation existing