PostgreSQL 12 High Availability Cookbook - Third Edition

There's more...

There are limited circumstances where a node cannot be recovered, rewound, or rejoined. In these instances, it becomes necessary to essentially rebuild the node. When this happens, we can basically follow the procedure outlined in the Cloning a database with repmgr recipe to reintroduce the node to the cluster. If we're brave, we can even truncate the procedure to these commands as the postgres user on the failed node:

sudo systemctl stop postgresql@12-main
rm -Rf /db/pgdata
repmgr standby clone -h pgha2 -d repmgr
sudo systemctl start postgresql@12-main

In past engagements, the author has even written scripts that attempt to use node rejoin, and if that command returns a nonzero exit code, it rebuilds the node, as shown ...