restart master first in some edge cases

[FxKu]

The current master branch of operator will always restart replicas first when attempting a restart if Patroni config differs from manifest. But in case the following parameters have lower values:

• max_connections
• max_prepared_transactions
• max_locks_per_transaction
• max_worker_processes
• max_wal_senders

master needs to be restarted first. Problem is SetConfig is only called once and there might be a mix of variables that require different restart strategies. One idea could be to first check if there's a pending restart on the master and restart it, if the mentioned variables have lower values.

Or we accept a mismatch between Patroni config and actual config since a restart is only triggered by the operator if manifest differs from the result of /config endpoint.

The Postgres version in the manifest will be raised back to 14 when the e2e test image is on 14 as well.

[FxKu]

👍

[Jan-M]

👍