2 Replies Latest reply on Sep 23, 2010 3:17 AM by Philip Bauwens

    Jboss clustering with multiple networks

    Philip Bauwens Newbie

      Hi,

       

      I've got two servers, both running jboss5.1.0.  I want to make one cluster of both machines.

      Both server belong to two networks, 10.255.245.x and to 10.255.212.x network.

       

      I start jboss with command ./run.sh -c SMMSvcTier -b 0.0.0.0 -Djboss.messaging.ServerPeerID=1 on the first machine.

      Next, I start jboss with command ./run.sh -c SMMSvcTier -b 0.0.0.0 -Djboss.messaging.ServerPeerID=2 on the 2nd machine.

       

      Log file does not show any errors, but the 2 cluster nodes do not seems to see each other.

       

       

      When I use -b option with value 10.255.212.50 for the 1st node and 10.255.212.52 for the 2nd node, both nodes see each other and the cluster is formed correctly.  The problem here is that I can no longer use my 10.255.245 network.

       

      I did not make any changes to the cluster configuration files as delivered with jboss.

       

      How can I make my cluster work while using 0.0.0.0 as bind address ?  Which files/parameters do I need to configure ?

        • 1. Re: Jboss clustering with multiple networks
          alex soto Newbie

          Philip,

           

          Try setting the system variable jgroups.udp.ip_ttl to a higher value.  The default is 2.

          This can also caused by the network switch dropping multicast packets when routing the traffic from one network to the next.

          Your best option is to keep both nodes in the same subnet.  Hope it helps.

          • 2. Re: Jboss clustering with multiple networks
            Philip Bauwens Newbie

            Alex,

             

            Thanks for your reply.  I had put aside the issue for some time due to other priorities.

            Your reply made me dig in to it again and I managed to solve the issue.

             

            After all, the problem was related to my /etc/hosts file.  The hostname was linked to the loopback adres, and apparently, the hostname is used for clustering config.  As a result, all clustering ports were listening on the loopback interface.