One way would be with 2 virtual IPs, round robin DNS, and heartbeat.
You configure heartbeat with 1 virtual ip with the 1st head node as primary and then the 2nd virtual IP on the 2nd head node as primary.
If one of the head nodes goes down then the other head node will take over the virtual IP and answer for both addresses until the down server is restored. Then simply add both virtual IPs to the DNS record so that each time a request for the domain is answered the IPs will be rotated in the answer and about 1/2 the traffic will go to each head node.
If you don't have 2 available IPs you can use heartbeat to set-up an active/backup solution using 1 IP and 1 active head node, with a live backup monitoring the primary/active head node.
Note, you will also have to deal with the private IP your nodes use as the gateway as part of the set-up.
This thread has a little more detail on this set-up: