2 Replies Latest reply on Sep 7, 2004 4:23 AM by citystory

    IP failover

    tperrigo

      We are currently configuring a couple of servers for IP failover-- if the primary machine goes down, the secondary will assume the same IP. My question is: how should we configure the jboss instance on the secondary machine? Should we cluster the two? How does clustering work with IP failover? As an experiment, I changed the IP of a machine with a running jboss server and UDP exceptions began occuring...Is there a way to configure jboss to handle havng the IP of the machine it's running on change out from underneath it? If so, I could just leave a server instance running on the secondary machine and allow the IP failover to happen.

      Any help or suggestions would be greatly appreciated!
      Tim

        • 1. Re: IP failover
          andyd

          Tim,
          I'm looking into a similar config. I believe the answer is to have fixed addresses for each server with another address which can be passed between them. I've just installed fake to try out. Ideally I'd like to write a singleton mbean to control handover of the address between nodes.

          Regards,
          Andy.

          • 2. Re: IP failover
            citystory

            How could this be done in Java? I'm interested.

            The ClusterAgent installed on each machine should be able to set the IP on a healthy or less loaded (detect this through heartbeat in their private network) to receive requests.

            Will JBoss Cluster implement it? I believe Oracle RAC or HP Service Guard should have implemented sth similar although not Java.