Ubuntu – Juju didn’t configure rabbitmq for openstack

cloudjujumaasopenstack

I have installed ubuntu Openstack HA with juju with all 24 servers. But my openstack is not working at all. On dashboard on every page I get errors saying "could not retrieve usage information", "could not retrieve volume information, "could not retrieve …..etc

I spent hours and have discovered that juju has not done configuration correctly. I found that on cloud controller in nova.conf juju has added rabbitmq vhost entry, but that virtual host is not added in rabbitmq. Then how is it supposed to work? And on juju-gui canvas rabbitmq is all green and is working fine, which in reality it is not.

I am really wondering if juju has really done correct configuration in all 24 servers now, I am getting the feeling that it would have been faster if I would have done openstack deployment manually instead of using juju. Why was the virtual host entry not added in rabbitmq? How should I solve this?

Best Answer

To resolved your immediate problem (lack of vhost creation in RabbitMQ - which is odd) I'd remove and then re-add the relation between nova-cloud-controller and RabbitMQ:

juju remove-relation rabbitmq-server nova-cloud-controller
juju add-relation rabbitmq-server nova-cloud-controller

This should force re-creation of the vhost in RabbitMQ; the RabbitMQ charm should have done this anyway so it would be good to dig into this further to figure out what went wrong; are any vhosts listed if you login to one of the rabbitmq service units and run:

sudo /usr/sbin/rabbitmqctl list_vhosts

The problem you are seeing would indicate a bug in the charm/some sort of race condition during cluster bring-up for RabbitMQ.

Related Question