PostgreSQL 12 High Availability Cookbook - Third Edition

How it works...

This recipe is one of the easiest in our list, but it is equally important. We begin by using drbdadm to disconnect each node from the communication link DRBD uses to copy data between servers.

Then, we use drbdadm again to doubly invalidate the data on the bad node. First, we use the invalidate-remote parameter on pg1 to ensure it sees pg2 as unusable. Then, we use the invalidate parameter on pg2, so it sees its own data as incorrect. We can see what this looks like by examining the contents of /proc/drbd again:

As you can see here, DRBD considers the data on the current node as Inconsistent, meaning it cannot be used as the ...