ironic-neutron-agent fails after enabling rabbitmq_quorum_queues in stable 2023.1

[Antony Cleave]

Hi

I have been experimenting with the new om_enable_rabbitmq_quorum_queues feature in stable/2023.1 (commit 3b2dc26)

The ansible run completes and I have run the process from the docs page to reset the rabbit state and restart the containers.

All containers seem to be happy except ironic-neutron-agent

this fails in a loop which you can see in the combined log file I have attached for the same time period for the ironic_neutron_agent container and the rabbitmq container it is communicating with. I think the fix will be to remove the auto_delete property from the queue when quorum mode is active but I'm not sure where to start looking.

I have tried this with a build of image 16.2.1 which I built some time before xmas and version stable/16.3.0 of the ironic-neutron-agent container built this afternoon and no change in behavior is noted. The logs are from 16.3.0 of the ironic agent and 16.2.1 of rabbitmq. I am pretty sure the error is on the Ironic agent side or all containers would be affected.

I am running a hyperconverged setup on ubuntu 22.04 with 3 controller/hypervisor nodes.

The containers were built in ubuntu22.04

Thanks in advance

Antony

[Antony Cleave (antony-cleave)]

applying this patch https://review.opendev.org/c/openstack/networking-baremetal/+/903995/1..2/networking_baremetal/agent/ironic_neutron_agent.py and resetting the rabbitmq state with

kolla-ansible rabbitmq-reset-state

fixes it