How HA software works (part 2)


System Monitoring is the key.

There are several ways that the cluster can monitor itself, but the most accepted is called heartbeating.
Essentially, the machines ping each other on a defined interval to continuously confirm that all machines are operational. Equate this to a heartbeat monitor.
In the event that the primary machine does not heartbeat back, the software begins a procedure to replace it with a secondary machine. this is known as a failover.
Most HA software solutions do a lot of error checking before actually failing over a server, to ensure that the heartbeat is not delayed or stopped by some other condition.