Scale environment 3 controllers, 45 computes Mitaka, L3 HA
After running rally test create_and_delete_routers agent logs are full of errors http://paste.openstack.org/show/525987/ and on server side http://paste.openstack.org/show/525175/.
The root cause of this issue seems that deletion of router is not isolated properly, HA ports are deleted concurrently, so we have HA router proceed, although HA interfaces are deleted.
In logs we can see "Got router deleted notification for <id> router_deleted" and after that still some actions with that router, warning "Info for router <id> was not found. Performing router cleanup" and after that "Got routers updated notification :[u'<id>'] routers_updated", all this caused infinitive loop for deleting routers until l3-agent restarted.
Logs - https://drive.google.com/a/mirantis.com/folderview?id=0By7EBG7Kg4gHXzJpRUNiS1MzU2M&usp=sharing#
The trace that Ann posted describes the agent-side logs (which are dealt with in https:/
Even though these 2 bugs describe a similar issue, we should use both to track the server-side and client-side bugs (the client-side bugs aren't limited to l3-ha).