It's time to address the elephant in the room. When managing a wide PostgreSQL cluster, we will often need to rebuild, reassign, and repair nodes that are replicas of our primary server. If we remember our rule-of-threes, three or more nodes make it difficult and error prone to perform any task related to replication.

While Barman and OmniPITR are useful, neither of them is capable of managing a wide network of PostgreSQL replicas. This is why we would like to thank 2ndQuadrant for repmgr. With it, we can create new clones and add them to an existing cluster of PostgreSQL servers. We can shut down the existing primary server and promote any node in this network. Further, all of the existing replicas automatically consider the promoted node their new source of streaming updates.

This may not be the first tool to perform this task, but it is one of the best available. We'll tackle the process of installing it in this recipe before moving on to usage scenarios...