Infrastructure roles and cluster pods restarts

[mirkomarchiori]

Hi, we are using the ConfigMap and Secret postgresql-infrastructure-roles combo to define some postgres roles that will be created for every database we deploy.

Here is an example of such resources:

Secret:

apiVersion: v1
kind: Secret
metadata:
  name: postgresql-infrastructure-roles
  namespace: pg-admin
stringData:
  monitoruser: <PASSWORD>

ConfigMap:

apiVersion: v1
kind: ConfigMap
metadata:
  name: postgresql-infrastructure-roles
  namespace: pg-admin
data:
  monitoruser: |
    inrole: [pg_read_all_stats]
    user_flags:
      - login

Then, those "global" users are allowed on the pg_hba configuration inside each cluster, in the kind: postgresql resource.

The operator is deployed through its Helm Chart on a specific namespace and is able to manage postgres clusters on all other namespaces.

Now, my question is: what happens when I add a new infrastructure role on the Secret and ConfigMap?

• Do I need to restart the operator for it to add those users on each downstream cluster namespace?
• Will the pods of all my already deployed databases be restarted?
• Does the enable_lazy_spilo_upgrade flag of the operator help in avoiding pods restarts in this case? (I am asking because this is not really a kubernetes change, but more of a postgres configuration I guess)

[mirkomarchiori]

Does someone know anything about this? I would like to know whether this roles update will trigger pod restarts or not.