PostgreSQL 12 High Availability Cookbook - Third Edition

There's more...

If we use firewall rules, we can even watch the fencing process in action. In our cluster, the IP addresses 10.0.30.2 and 10.0.30.3 correspond to the standby and witness respectively. We can block all traffic from those nodes with these commands:

sudo iptables -A OUTPUT -s 10.0.30.2/32 -j REJECT
sudo iptables -A OUTPUT -s 10.0.30.3/32 -j REJECT
sudo iptables -A INPUT -s 10.0.30.2/32 -j REJECT
sudo iptables -A INPUT -s 10.0.30.3/32 -j REJECT

Then we would want to observe the primary's reaction by watching the log like this:

tail -f /var/log/postgresql/repmgr.log

After a few seconds, the primary will notice that communication to both the standby and witness is no longer possible. This is what the logs look will like at this ...