neutron: restart neutron-ha-tool when the config file changes

[dirkmueller]

When we toggle ssl/non-ssl in keystone, neutron-ha-tool
gets a new config file but doesn't get restarted, so it just
repeatedly crashes all the way until it causes a pacemaker
failcount exceeded and then its dead. We should try better.

[dirkmueller]

(needs testing, and probably does not work as is)

[aspiers]

LGTM. CI failure is unrelated, something to do with missing iso?

[dirkmueller]

yeah, not sure about that problem, it happens quite frequently recently.

[cmurphy]

I would like to see the HA job passing as this is supposed to be fixing that job.

[cmurphy]

I don't think this will work, see inline comment

Looks good to me now, will see what CI thinks

[sjamgade]

Error executing action `restart` on resource 'pacemaker_primitive[neutron-l3-ha-service]' from CI logs

[dirkmueller]

yes, thats why there is "do not merge" and "wip" and "untested, probably does not work" on the PR

[vuntz]

Ah, I hadn't realized. In case it's not already solved on your hard disk: the restart action needs to be on a service resource, not a pacemaker resource. So we need to define a service for neutron-l3-ha-service, make sure the service uses the pacemaker provider, and that should do it.

[dirkmueller]

@vuntz thanks, that did it