ipfail is very simple to configure.
First, you must tell the HeartbeatProgram to start ipfail. This is done with the respawn directive. Add the following line to your ha.cf file:
respawn hacluster /usr/lib/heartbeat/ipfail
Note: Do not start ipfail as root. While the LinuxHAckers are not aware of any dangerous code in ipfail, it is safest to run as little code as possible with root privileges. ipfail has absolutely no need to run as root, so do not expose yourself unnecessarily to a dangerous configuration.
Next, you must select and configure some PingNodes. The selection of a strategic PingNode (or ping_group) will vary between each installation, but it is usually a good idea to select a pingable switch or router. Consider that there are many criteria involved in selecting a proper device to ping. If you select a router that is 13 hops away, you may not be able to infer anything when the node dies. Your connectivity could still be OK. Likewise, selection the very next hop might not reveal enough about the state of the network. Certainly, when the closest router is unreachable, you have a problem, but if it is still reachable even though the gateway beyond it is not, your users may not be able to get to the services you are providing. This part of the configuration is a bit of an art, but you can boil it down to logical conclusions when you ask the question: What do I know if this ping node suddenly goes unreachable?
Also, consider whether the owner of the device that you plan on pinging will mind your ICMP echos. Heartbeat will send out a ping every keepalive seconds. If the device will drop pings under heavy load it is also a poor choice. You essentially want to pick solid equipment with a high MTBF which is nearly always available (preferably even more available than your cluster) and which will not generate complaints by the network administrator.
To add a PingNode, use a PingDirective.
Finally, start heartbeat and ipfail will be active.