I have two app servers that will be in a JBoss cluster. I want to use multicast for auto discovery. However, my client does not want to pay for an extra machine (or HW load balancer for that matter) for load balancing/redundancy. Therefore I came up with a utility almost like "fake", which takes over the IP address on one of the NIC's (both are multihomed) of the failing node - in order for clients (which are partitioned) to fail over to the node still running.
I know this is crude but they find this acceptable. My problem is that I have grave reservations that me manipulating the IP addresses of the nodes like that could cause the multicast/clustering of JBoss to fail. If one node is down, the other node will have three IP adresses - the internal IP used on one interface, then the node's own IP on the other interface as well as the taken over IP of the other node.
Will this work? If not, how should I go about to handle this scenario?
Why don't you simply try and give us some feedback?