Rabbitmq CPU consumption increase after upgrade to version 3.11.18

[briceviard]

Hi,
We upgraded our rabbitmq cluster from version 3.11.9 to version 3.11.18 and we noticed the CPU consumption increased ( + 30%).
We had to increase cpu capacity to avoid trottling :-(

I didn't find anything about that in release note, if someone has an explanation?
Can we tune something (rabbitmq or erlang paramters) to optimize that....?

Thanks in advance.
Regards, Brice.

[michaelklishin]

We don't have enough information here to suggest anything specific.

See

• Thread Statistics: How is Scheduler and CPU Time Spent?
• CPU utilisation
• Reducing CPU Usage for "Moderately Idle" Nodes and Clusters

Something about your workload or connection, queue or stream leader distribution might have changed around the same time. For example, did you rabbitmq-queues rebalance queue leaders after all nodes were restarted? If not they may be residing on the same node, which means that node is guaranteed to use more CPU resources.

[briceviard]

I think I found what was wrong :-(

We used command "rabbitmqctl status" in our livenessProbe and readinessProbe to check that rabbitmq is working fine...
I changed this to only check the tcp port and the cpu consumption descrease :

Do you have recommendation for readinessProbe and livenessProbe to check rabbitmq?
Thanks.

[lukebakken]

https://www.rabbitmq.com/monitoring.html

https://www.rabbitmq.com/monitoring.html#health-checks

[mkuratczyk]

I think we should update these docs. Our operator uses a TCP check for a readinessProbe and doesn't use livenessProbe at all. This is the best recommendation we have.

[mkuratczyk]

well, using the Operator is another recommendation, so that you can avoid problems and won't need to ask for our recommendations, because they are already baked-in :)

[briceviard]

ok, thanks for your help :-)