No promotion to new primary of newer instances

[ExtremTrucker]

We are currently attempting to modify the PVC in a Postgres cluster (3 replicas) using the "Resize PVCs With StorageClass That Does Not Allow Expansion" method.

We have enabled synchronous replication and failsafe mode for the cluster.

  patroni:
    dynamicConfiguration:
      failsafe_mode: true
      synchronous_mode: false
      synchronous_mode_strict: false
      postgresql:
        parameters:
          synchronous_commit: 'on'

Once the second instances have been fully booted up, the first instance is removed from the cluster and no new master is designated.

If the Patroni section is not active, everything works as expected, but unfortunately that is not a solution, as data integrity is compromised.

Are we overlooking a necessary step in our case? Or is it unfortunately necessary to disable the Patroni option?

A switch from the primary replica to the second instance was unsuccessful because the switchover did not complete.

[ExtremTrucker]

One way to work around the problem without turning off the synchronous replica is to scale the replica down to 1 in the first instances so that the synchronous replica is one from the new instance pool.

This ensures that the new master is determined correctly right away without causing any problems with the quorum.

Is this the right approach?