List per-queue memory usage and number of messages:
rabbitmqctl list_queues name memory messages
Minor versions can be upgraded one node at a time.
Whole cluster has to be stopped for the upgrade.
Proposed HA for this upgrade (if possible):
- spin up temporary instance or cluster
- move producers to temporary cluster
- let consumers empty all queues from original cluster
- move consumers to temporary cluster
- upgrade original cluster
- move producers back to original cluster
- let consumers empty all queues from temporary cluster
- move consumers to original cluster
Note: start disk nodes first. Any RAM node will emit errors if there are no disk nodes present in the cluster.
Unless configured otherwise, following ports need to be open among cluster members:
- TCP/4369
- TCP/25672