Improve procedure for cluster restart to include particular case
There is a pitfall in the restart procedure. The problem regards all those parameters that can not be changed to a lower value in replicas while the primary maintain a higher value. In such case a switchover is not possible and a restart of the primary is required.