'openstack-nova-novncproxy-clone' has 'target-role=stopped' after Pacemaker restart and fencing, then will not restart via Pacemaker.

Issue

Restarting openstack-nova-novncproxy-clone using the following pcs command causes Pacemaker ops_params to timeout after 20 seconds, resulting in the respective OpenStack Platform controller being fenced:

pcs resource restart openstack-nova-novncproxy-clone

After the restart is initiated, logging in /var/log/messages shows systemd attempting to stop openstack-nova-novncproxy:

systemd: Stopping Cluster Controlled openstack-nova-novncproxy...

If that request continues for 20 seconds, the following lrmd and crmd errors are logged also in /var/log/messages, indicating that the 'ops_params' Pacemaker service resource stop timeout has been triggered, which fences the controller node:

lrmd[<PID>]: notice: Giving up on openstack-nova-novncproxy stop (rc=196): timeout (elapsed=19987ms, remaining=13ms)
crmd[<PID>]: error: Operation openstack-nova-novncproxy_stop_0: Timed Out (node=<controller_name>, call=759, timeout=20000ms)

Once the controller has restarted, openstack-nova-novncproxy-clone does not restart successfully on boot under Pacemaker and its target role is target-role=stopped.