[library] fixing notifications.info rabbitmq when adding in a second external network

[Gil McGrath]

After completing a FUEL 4.1 openstack deployment using multi--node (NO HA) I needed to add in an 2nd external network.
Everything functionally worked - but I got the following error in /var/log/rabbitmq:

channel 1 - error: {amqp_error,precondition_failed, "inequivalent arg 'x-ha-policy'for queue 'notifications.info' in vhost '/': received the value 'all' of type 'longstr' but current is none", 'queue.declare'}

To fix this on my fuel server I edited the file: /etc/puppet/modules/neutron/manifests/init.pp
I edited line 77 to: 'DEFAULT/rabbit_ha_queues': value => 'False';

After this I deployed my openstack environment and went ahead and added in my second network. This seemed to fix my problem.

This is certainly not the correct fix, but it highlights the issue I encountered.

You may not consider this an real FUEL bug since I went ahead an modified my openstack environment post FUEL deployment.
But I thought I would report my issue.

[Bogdan Dobrelya (bogdando)]

AFAIK, rabbit_ha_queues should never be used for non HA multinode.

[Dmitry Borodaenko (angdraug)]

Line 77 in the 4.1 version of neutron/manifests/init.pp is inside the true branch of the "if $neutron_config['amqp']['ha_mode']" conditional, implying that sanitize_neutron_config has generated the wrong value for ha_mode field in non-HA deployment mode. Blueprint https://blueprints.launchpad.net/fuel/+spec/ml2-neutron targeted for 5.1 will get rid of this code, so this bug will disappear on its own in 5.1, and there's no point in fixing it in previous releases.