There is regression in EAP 7.1/Artemis HA caused by change in behaviour of NFSv4 between RHEL 7.3 and 7.4.

Customer impact: Artemis is not be able to guarantee HA on NFSv4 with RHEL 7.4. To recover from previous failure manual intervention and unavailable of customer service is required. There is no workaround.

Description of problem:
If Artemis is using NFSv4 as shared store for its messaging journal to provide HA then write to server.lock file made by live server is not propagated to server.lock file to backup server. This prevents failback to happen. This is change in behaviour in data synchronization between two NFSv4 clients between RHEL 7.3 and RHEL 7.4.

Mount options (taken from /etc/fstab):
10.40.4.25:/mnt/shared /mnt/hornetq/client nfs4 minorversion=0,rw,sync,nosuid,nodev,intr,tcp,noac,soft,lookupcache=none 0 0

Test scenario
Start 2 EAP 7.1 servers on different RHEL 7.4 machines using NFSv4 shared store
Kill EAP 7.1 with Artemis live and wait for failover
Start EAP 7.1 with Artemis live

Actual results:
Backup does not detect change in server.lock file made by live and failback does not happen.

Expected results:
Backup will detect changes in server.lock file and will shutdown. Live will activate and failback succeeds.

Steps to Reproduce:

• On 2 different RHEL 7.4 machines mount NFSv4 directory using mount options: minorversion=0,rw,sync,nosuid,nodev,intr,tcp,noac,soft,lookupcache=none
• In mount directory run backup.groovy script: "groovy backup.groovy" on first machine
• In mount directory run live.groovy script by groovy live.groovy on second machine