Hi,
I am investigating how to set up a cluster of wildfly (active) - wildfly (active) configuration with each wildfly 10 instance having a live-backup pair of message servers. The approach is data replication.
I followed the document to set it up the cluster
Chapter 29. High Availability - Red Hat Customer Portal
The configuration did not work completely since some up-processed messages of the wildfly instance which was shut down were not accessible to other wildfly instance to get processed. I made a simple modification to add discovery-group in the cluster-connection definition in the backup server, and I was able to observe message fail over to another wildfly instance (got processed). That sounds all good.
Test:
wildfly #1 (live1, backup2)
wildfly #2 (live2, backup1)
However, this change brought in an issue: the wildfly instance (wildfly #1) that was down could not restart, and it is complaining about the missing dependency of a queue. I can fix the issue by shutting down the other wildfly instance (wildfly #2) before re-starting the instance that was brought down. However, this is not a good solution in production.
If I shut down wildfly #2, I can restart wildfly #1 without problem.
Is it possible that the backup server (backup1) in the other instance (wildfly #2) was not shut down once wildfly #1 was restarted? The default value of allow-failback in the backup server is true, so there is no need to set it true.
Is there any other configuration that can get the wildfly #1 to restart without problem?
Thanks,
Wayne