rabbit_ha_queues parameter doesn't enabled for HA deployments

[Timur Nurlygayanov]

We use Mirantis OpenStack with the upstream puppet manifests for Heat deployment and we can see that in HA deployments we have the following parameter in Heat configuration file:

rabbit_ha_queues=False

In the result, if we will shutdown primary controller Heat will lost some queues and will doesn't work, we need to restart all Heat services to fix the issue. We investigated the issue and looks like all other services have correct configuration with rabbit_ha_queues=True, but Heat has rabbit_ha_queues=False.

Here is the MOS bug about this issue: https://bugs.launchpad.net/fuel/+bug/1465840

We tested Heat with rabbit_ha_queues=True, which we applied manually, it works well.

We compared puppet manifests for Keystone and Heat and found the differences:

Keystone:
https://github.com/openstack/puppet-keystone/blob/bca04a399e6de63bd6d3d3d985aeaf3134802ac1/manifests/init.pp#L764-772

Heat:
https://github.com/openstack/puppet-keystone/blob/bca04a399e6de63bd6d3d3d985aeaf3134802ac1/manifests/init.pp#L764-772

[Timur Nurlygayanov (tnurlygayanov)]

Fix is on review:
https://review.openstack.org/#/c/193492/

[Bogdan Dobrelya (bogdando)]

AFAIK, there is no sense to set up ha queues configuration for clustersize of 1

[OpenStack Infra (hudson-openstack)]

Change abandoned by Timur Nurlygayanov (<email address hidden>) on branch: master
Review: https://review.openstack.org/193492
Reason: We found the root of this issue, we fixed it with another commit: https://review.openstack.org/#/c/193511/

[Matt Fischer (mfisch)]

Bogdan, yes there is. If you are talking to rabbit via a proxy. We use it all the time.