9 Replies Latest reply on Oct 7, 2002 3:33 AM by cosmotic

    JBoss/Tomcat with JDK 1.4.0/1.4.1 on Linux cluster hangs at

      [I've posted this in the Installation Forums as well.]

      I'm trying to run JBoss-3.0.2_Tomcat-4.0.4 with JDK 1.4.0_02 and 1.4.1 on SuSE Linux 8.0. However, the startup hangs:

      INFO [ClusterPartition] Starting
      INFO [ClusterPartition] Connecting to channel

      Absolutely no warnings or errors, just nothing more.

      Otoh, it works fine with JDK 1.3.1:

      INFO [ClusterPartition] Starting
      INFO [ClusterPartition] Connecting to channel
      INFO [STDOUT]
      -------------------------------------------------------
      GMS: address is linux:32861
      -------------------------------------------------------
      INFO [DefaultPartition] New cluster view: 0 ([linux:32861])

      What could be the problem? Is JavaGroups Multicast incompatible in some way? Any hints?

        • 1. Re: JBoss/Tomcat with JDK 1.4.0/1.4.1 on Linux cluster hangs
          belaban

          Can you enable the full logs for JavaGroups and send them to me (belaban@yahoo.com) directly ? You do this by creating a file called javagroups.properties and placing it either in your home directory or somewhere on the CLASSPATH. The contents should be

          trace=true
          timestamp_format=HH:mm:ss[SSS]
          default_output=WARN /tmp/trace.all


          Change the /tmp/trace.all to c:\trace.all if you're on Windows.
          Cheers,.
          Bela

          • 2. Re: JBoss/Tomcat with JDK 1.4.0/1.4.1 on Linux cluster hangs

            I couldn't get tracing to work yet, but log4j gives the following output:

            INFO [ClusterPartition] Starting
            INFO [ClusterPartition] Connecting to channel
            INFO [startWork()] creating sockets and starting threads
            INFO [createSockets()] all sockets will use interface 127.0.0.2
            FATAL [up()] exception creating sockets: java.net.SocketException: bad argument for IP_MULTICAST_IF: address not bound to any interface

            • 3. Re: JBoss/Tomcat with JDK 1.4.0/1.4.1 on Linux cluster hangs

              With JDK 1.3.1, it looks like this:

              INFO [ClusterPartition] Starting
              INFO [ClusterPartition] Connecting to channel
              INFO [startWork()] creating sockets and starting threads
              INFO [createSockets()] all sockets will use interface 127.0.0.2
              INFO [createSockets()] socket information:
              local_addr=linux:32795, mcast_addr=228.1.2.3:45566, ttl=64
              send socket: bound to 127.0.0.2:32794, send buffer size=64000
              receive socket: bound to 127.0.0.2:32795, receive buffer size=128000
              multicast socket: bound to 127.0.0.2:45566, send buffer size=131070, receive buffer size=131070
              INFO [startThreads()] created unicast receiver thread
              INFO [STDOUT]
              -------------------------------------------------------
              GMS: address is linux:32795
              -------------------------------------------------------

              • 4. Re: JBoss/Tomcat with JDK 1.4.0/1.4.1 on Linux cluster hangs
                belaban

                What address is 127.0.0.2 ? It should be 127.0.0.1 (loopback) !

                Check cluster-service.xml for the correct spec. There should be something like

                "UDP(mcast_addr=228.1.2.3;mcast_port=45566):" +
                "PING:" +
                "FD(timeout=5000):" +
                "VERIFY_SUSPECT(timeout=1500):" +
                "MERGE:" +
                "NAKACK:" +
                ...


                Check whether bind_addr is present. If yes, remove it. If not, set it to (e.g. 127.0.0.1).

                UDP(bind_addr=127.0.0.1;...)

                Note that this will only allow your nodes to run on the same machine. If you want to run on different boxes, bind_addr needs to be bound to (e.g. eth0), or some other routable network interface which can reach the other box.

                Bela

                • 5. Re: JBoss/Tomcat with JDK 1.4.0/1.4.1 on Linux cluster hangs

                  Still odd that it works on JDK 1.3.1 with the exact same settings.

                  Anyway, I'll try to change the bind address then - on Monday. ;)

                  • 6. Re: JBoss/Tomcat with JDK 1.4.0/1.4.1 on Linux cluster hangs

                    OK, it's working now. Thanks for the help.

                    I hadn't configured the protocol stack, so after adding the default string from the Clustering docs and setting bind_addr to the machine's IP, it began working on JDK 1.4.1 too.

                    Yet I don't have a clue where the 127.0.0.2 comes from and why it works with JDK 1.3 but not 1.4...

                    • 7. Re: JBoss/Tomcat with JDK 1.4.0/1.4.1 on Linux cluster hangs

                      Ah, wait, in /etc/hosts:

                      127.0.0.2 linux

                      Anyway, this is a Linux (or SuSE) specific issue, so I better ask elsewhere. ;)

                      • 8. Re: JBoss/Tomcat with JDK 1.4.0/1.4.1 on Linux cluster hangs
                        belaban

                        > Ah, wait, in /etc/hosts:
                        >
                        > 127.0.0.2 linux

                        Remove this from /etc/hosts, add something like (if not yet there):
                        127.0.0.1 localhost

                        • 9. Re: JBoss/Tomcat with JDK 1.4.0/1.4.1 on Linux cluster hangs

                          Yes, after commenting it out, everything works fine. Seems to be a little bug in the auto-config of SuSE. Thanks for your help. :)