The current RabbitMQ cluster isn't actually clustered. This hasn't been a problem since everyone was connecting to rabbitmq01 and the cluster wasn't behind a load balancer. Now that it is, clients that get bounced to rabbitmq02 or 03 get rejected because they aren't part of the cluster. I think the problem is that when they were first built they were not configured to cluster and once they start they won't auto-discover their peers.
While I can use the CLI to manually configure the cluster, I'd prefer these two nodes to be rebuilt to ensure the ansible playbook and role configuration is correct on fresh nodes now.
Sooner is better as applications are starting to move to RabbitMQ.
Metadata Update from @kevin:
- Issue close_status updated to: Fixed
- Issue status updated to: Closed (was: Open)
to comment on this ticket.