neutron-server process get restarted when shared-db-relation-changed occurs

[Yoshi Kadokawa]

Whenever shared-db-relation-changed has been triggered,
for example leader of percona-cluster is changed, neutron-server process in neutron-api application, will get restarted.
percona-cluster is in HA with hacluster charm, so even one of the percona-cluster unit is being failed over,
the VIP will just go to another unit and nothing should affect to other services,
so restarting services should not be required.

I am seeing this when doing the following steps.

OS: Ubuntu 16.04
OpenStack Mitaka

1. Deploy OpenStack with this bundle[0]
2. Shutoff percona-cluster unit who is the leader
3. Confirmed that neutron-server process got restarted in neutorn-api unit

[0] http://paste.ubuntu.com/25233464/

[Edward Hope-Morley (hopem)]

confirmed.

[Edward Hope-Morley (hopem)]

The cause of this appears to simply be that the neutron-api charm will always (unless the unit is paused) do a db init - https://github.com/openstack/charm-neutron-api/blob/master/hooks/neutron_api_hooks.py#L364

We should implement the db init check that already exists in the cinder charm to protect against this.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/493931

[Edward Hope-Morley (hopem)]

neutron-api patch actually landed already and was released in 17.08