As an intro to clustering - an active/passive demo:

1 - set up 2 nodes with matching webserver contents (pop in Knoppix CD)
2 - have one student log into the virtual ip with telnet another with http
telnet connection should verify he is actually on node 1 (use hostname -a )
3 - fail node 1 (power off)

the telnet session should die, the web surfer should not notice any change.
Telnet back into the virtual ip. Telnet session will now be running on node 2.
You have just demonstrated a sucessful fail over in a HA enviorment.
Class, Please notice the lack of impact on a web connection when hardware dies.

Topics for lecture - High availability, session oriented vs connection oriented protocols, recovery concerns. gratuitious ARP.

for next class - Fail over (active/passive) vs distributed workload (active/active) clustering. Study notes on Mosix and Beowulf

Demonstration outlined above was done after we installed heartbeat on two RH servers at work. We used a crossover cable for the cluster connection, but for a classroom demo, you could use a single network cable for cluster and network. The virtual ip is transferred via "gratuitious ARP" where the backup refreshes the ARP cache with his own physical address in response to a failure of the primary.