Created attachment 943010 [details] output of rabbitmqctl list_queues name messages_ready messages_unacknowledged consumers auto_delete memory Description of problem: Running in the scale lab, the rabbitmq server is using a lot of memory that we can't find a reason for Version-Release number of selected component (if applicable): rabbitmq-server.noarch 3.1.5-6.3.el7ost How reproducible: Every time Steps to Reproduce: 1. Run rabbitmq at scale 2. 3. Actual results: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 3671 rabbitmq 20 0 16.748g 0.012t 2320 S 57.1 10.1 5195:41 beam.smp Expected results: Additional info: I have attached a file with the output of `rabbitmqctl list_queues name messages_ready messages_unacknowledged consumers auto_delete memory`
This looks very much like the issue with oslo.messaging we've fixed already (or going to fix soon). Unfortunately the ticket is more that 1 year old, however I believe we can collect here some interesting tips for further investigation of a similar issues. * A queue "notifications.info" isn't empty. This might mean that ceilometer-agent is not running. One should either delete this queue (setting "notification_driver = noop") or start ceilometer-agent. Another one possible consumer of the messages in this queue is CloudForms - either ensure it's actually running and consuming messages, or delete the queue. * There are lots of queues "reply_<something>" w/o any alive reader or with full of messages. This looks like the issue with restarted clients and their autodeleted queues which were improperly declared. THis was fixed in python-oslo-messaging-1.3.1-2.g33f3b18.el7ost. Also this patch is strongly required, however the package wasn't officially built yet - https://code.engineering.redhat.com/gerrit/#/c/60809/ . Thus I'm closing this as a duplicate of the bug 1262455. Feel free to reopen if this is a different issue. *** This bug has been marked as a duplicate of bug 1262455 ***