Comment 26 for bug 1927868

Corey, Jared I believe your analysis is running a little in the wrong direction here:

1) We run OpenStack TRAIN (15) and also experienced the described issues. So there cannot be any relation to the database schema upgrades.

2) We did experience the issue even before the recent upgrade and we believe it's a coincidence that it was triggered just now. There also is a bug report about exactly the observations I made (all gateway interfaces where down): https://bugs.launchpad.net/neutron/+bug/1916024
Did you observe anything else while running the potentially flawed 16.3.1 and when looking the interfaces of a virtual router?

3) The commit https://opendev.org/openstack/neutron/commit/12c07ba3ea9c6501dd7494561e2920496407c48b was just about trying to fight a race condition with a static "sleep 3", but it also gave us an error message now and we saw this multiple times during a larger batch of machine migrations. So the issue is real in TRAIN and likely also in USSURI and later.