I am trying to figure out a clear way to preform a post-disaster (where the master hasn't had the chance to shut down properly) master failover in a multi-standby streaming replication setup. I found a couple of old posts suggesting that the choice of the new master is very important and if not done correctly might lead to database corruption.The posts suggest that the first thing to do is to find the most caught-up replica, by compare WAL replay locations among all standby servers. This makes perfect scene but also makes the failover process a little bit more complicated and difficult to automate. I guess my question is, does my understanding is right and if there are any changes in this regard in the newer Postgresql 9.3/9.4 releases.
