Again diskspace issues on the testpod. notifications.sample queue was very large (24G). Threw away huge directory containing quorum Q data.
Threw away all mysql backups.
Rebooted lxchost1 tesptod.
Noticed ceilometer also had huge logs.
Purged notifications.info queue as I saw alot of spam in ceilometer-agent-notification.log on ceilometer node about duplicate messages.
Also noticed unread messages in cinder-volume queue.
It shows INTERNAL_ERRORs and failure to connect with rabbitmq. 
Stopped the cinder-volume service.
Purged cinder-volume queue.
Started ceilometer-agent-notification
Right away seeing the same INTERNAL_ERROR messages as on cinder before.