Comment 45 for bug 1818113

@aurelion-lourot

Not sure if this is related to your fix, but in https://solutions.qa.canonical.com/qa/testRun/5589fe2d-bac5-4f35-80a1-57f4a3e454c4 it took the keystone charm 45 minutes to recognize that percona was clustered and ready before starting any endpoint updates (those updates then ran for 1.5hr+ before the run timed out). This is not behavior we have previously seen from keystone before using your fork. This seems like an inordinate amount of time for a deployment which already takes 5+ hours.

/var/log/juju/unit-mysql-1.log
...
2020-07-15 20:29:50 DEBUG juju-log cluster:32: All 3 percona units reporting clustered
...

/var/log/juju/unit-keystone-1.log
...
2020-07-15 21:19:02 DEBUG juju-log identity-service:144: Not at expected scale, not enough units on shared-db relation
...
...
2020-07-15 21:23:12 INFO juju-log shared-db:80: Updating endpoint for gnocchi
...