Bug 1444143

Description of problem:

Rebooting octavia controller nodes breaks the configuration as the approach used in devstack and the current version of the tripleo deployment configuration does not survive the neutron_ovs_cleanup process on node bootup. 

Version-Release number of selected component (if applicable):

OSP 11 Beta, upstream pike

How reproducible:

100%

Steps to Reproduce:
Configure a working octavia deployment using the steps used in devstack or the the post deployment steps being developed here: 

https://review.openstack.org/#/c/447496/

Actual results:

Lack of communication with existing amphorae and unable to create new ones after reboot.

Expected results:

Normal operations should resume after rebooting.

Additional info:

Including all of the required details in interface definition files does *not* solve the issue as the interface is brought up on network start but is then deleted by neutron_ovs_cleanup. A cleaner solution might be to work out how octavia could plug the port itself similar to way dhcp and l3 agents do.

(NOTE: I'm initially reporting this against octavia, not osp-director as: a. it is suspect that there is no documented approach to creating a real-world deployment that would survive reboots in the Octavia documentation and b. I suspect that a good solution will probably require some kind of change in how octavia deals with these ports)