db-initialized is cleared when removing a unit

[Felipe Reyes]

When scaling in keystone and the unit removed is the leader
[Test Case]

juju deploy ./bundle.yaml
juju wait
juju status # identify the keystone leader unit
juju remove-unit keystone/N # N being the leader.

Expected result:

keystone stays running .

Actual result:

running keystone units get into blocked state with the message "Database not initialised"

keystone/0 blocked idle 3 10.5.2.10 5000/tcp Database not initialised
keystone/4* blocked idle 7 10.5.3.150 5000/tcp Database not initialised

Related bug 1797229
Related commit https://opendev.org/openstack/charm-keystone/commit/a87b1011066650dbfc924501202d7048f6f022bc

----- bundle.yaml ----
series: focal
applications:
  hacluster:
    charm: ch:hacluster
  keystone:
    charm: keystone
    channel: edge
    revision: 570
    num_units: 3
    options:
      openstack-origin: yoga
      token-expiration: 60
      vip: 10.5.250.120
  keystone-mysql-router:
    charm: mysql-router
    channel: edge
    revision: 27
  mysql-innodb-cluster:
    charm: mysql-innodb-cluster
    channel: edge
    revision: 25
    num_units: 3
    options:
      source: distro
relations:
- - keystone:shared-db
  - keystone-mysql-router:shared-db
- - hacluster:ha
  - keystone:ha
- - keystone-mysql-router:db-router
  - mysql-innodb-cluster:db-router

--------------------

[Billy Olsen (billy-olsen)]

Is this related to https://bugs.launchpad.net/charm-keystone/+bug/1942289 ? Seems like a duplicate to me?

[Felipe Reyes (freyes)]

it is, thanks ;-)