Restarting neutron-server can cause rabbitmq errors

[Thomas Bachman]

After deploying an ACI integration with the L3 ASR plugin via Mirantis OpenStack (MOS) 9, the deploy succeeds, but when using Horizon, there are errors and timeouts. The logs show things like this:

http://pastebin.com/FGzBKRg8

Specifically, this warning message:

Process forked after connection established! This can result in unpredictable behavior. See: http://docs.openstack.org/developer/oslo.messaging/transport.html

When this happens, restarting rabbitmq-server fixes it.

I don't see this error message when using the same installer to deploy with the APIC L3 plugin.