Description of problem: If reconnection occurs between openstack components and rabbitmq, for example, triggered by the server reboot, there are some queues in rabbitmq become orphaned (a.k.a., no consumer consume them), but they still accumulate RPC or fanout messages, which cause possible memory pressure. There are upstream fix to address this issue by setting TTL for direct and fanout queues instead of relying on auto_delete flag: https://review.openstack.org/#/c/243845/ (it's has been backported to liberty: https://review.openstack.org/#/c/279561/) https://review.openstack.org/#/c/292313/ The last patch is the latest, which changes the default value of rabbit_transient_queues_ttl to 30mins. Version-Release number of selected component (if applicable): python-oslo-messaging-2.5.0-4.el7ost.src.rpm How reproducible: Give rabbitmq node A has most master queues running, power off or restart the server where rabbitmq node A runs. Or following the reproduce steps in following bug reports: https://bugs.launchpad.net/oslo.messaging/+bug/1515278 or https://bugs.launchpad.net/oslo.messaging/+bug/1495568 Steps to Reproduce: 1. 2. 3. Actual results: Orphaned queues generated in rabbitmq, which can be verified by running 'rabbitmqctl list_queues name messages consumers' command. Expected results: no orphaned queues after a short period. Additional info: Related bug report in upstream: https://bugs.launchpad.net/oslo.messaging/+bug/1515278 https://bugs.launchpad.net/oslo.messaging/+bug/1495568 https://bugs.launchpad.net/oslo.messaging/+bug/1545002
*** This bug has been marked as a duplicate of bug 1310157 ***