Any time one primary PostgreSQL node is superseded by another, either because of a crash or planned switchover, it becomes necessary to return the node to the cluster as a standby. This returns the cluster to a fully symmetric pair so that the node can operate as a candidate in future migration events.
Rather than manually using pg_rewind, rsync, or some other intervention method, repmgr has a dedicated command that should be sufficient for returning the cluster to fully operational status. This recipe will explain how to return a node to the cluster and what limitations there might be while doing so.
