Coming from the shared LVM world where 'vgchange' is used to setup VGs to be used on each individual node (ie, lockstart, lockstop, activation, deactivation), it wasn't obvious that I didn't also need to run 'vgchange --persist y VG' on each and every node in the cluster.

The man page does state that it's stored in the VG metadata, but that could be a per node metadata item. It didn't help when I got stuck seeing the 'failed locking error' message as detailed in https://issues.redhat.com/browse/RHEL-137796 and I feared I was in some chicken and egg scenario, when in reality, I only needed to run '--persist start'.

In the man page's defense, the only shared example does properly start with the '--persist start' after the host_id, but when I was attempting the pr_key in lvmlocal scenario, I erroneously thought that the --setpersist y was required on all nodes.