Comment 4 for bug 1406629

Thanks for tackling this issue. I did test it based on a oslo.messaging 1.5.1 plus the preview patch of https://review.openstack.org/145232 and the issue worsened. After I did patch the impl_rabbit script and restart neutron-agent services for instance, the channels immediately increased without having even an rabbitmq failure during that test period. The rabbitmq-cluster was intact during this test.