Comment 3 for bug 1695662

Release-notes:

In Ubuntu 16.04.2 Newton cluster, following rabbitmq issue is seen sometimes.

http://rabbitmq.1065348.n5.nabble.com/Rabbitmq-boot-failure-with-quot-tables-not-present-quot-td24494.html#a24512

Workaround:

1) stop the rabbitmq on the node(service stop and epmd -kill) which is not clustered
(in a three node setup, mostly two of them were clustered together and the third one will be alone. it can also happen that all the rabbitmq nodes are alone too)
2) remove the /var/lib/rabbitmq/mnesia directory…
3) re-start the rabbitmq on the node.
4) check nova service-list output and restart all the services which are marked down. (may be unnecessary but found that nova-compute didn’t reregister if the rabbitmq went down. need to check it out later)