0 Replies Latest reply on Feb 17, 2011 5:16 AM by kakkarj

    Jboss Clustering: Issues with Unicast and performance

    kakkarj

      Hi All,

       

      We have 5 jboss nodes running in a clustering. I have changed the configuration to use the unicast model instead of multicast.

       

      I am facing few problems. Any help/comments would be much appreciated.

       

      Changed made to migrate to unicast from multicast :

      Switch all nodes to use a tcp-based jgroups stack   modified ..../bin/run.sh -c all -Djboss.default.jgroups.stack=tcp

       

      Updated the jgroups configuration and replace MPING with TCPPING and define the cluster members.  The relevant file is: $JBOSS_HOME/server/all/deploy/cluster/jgroups-channelfactory.sar/META-INF/jgroups-channelfactory-stacks.xml (File attached)

       

      Issues :

       

      1. Which model unicast vs multicast is better in terms of performance ?
      2. I see lots of connectiontable error message "2011-02-08 11:43:24,368 ERROR [org.jgroups.blocks.ConnectionTable] failed sending data to 10.240.2.33:48610: java.net.SocketException: Socket closed". Any idea why I get these error messages ?
      3. The logs shows that all my jboss nodes and postoffice can join the cluster on port 48610 however the logs below in red shows DistributedReplicantManagerImpl uses port 1599. Is this expected or am I missing any configuration here ?
        1. I commented the MPING block and added the following TCPPING block. What's the recommended configuration for port_range and num_initial_members considering I have 10 jboss nodes within the cluster.

       

            <stack name="tcp"

                ...............

       

       

                  <TCPPING timeout="3000"

                           initial_hosts="${jgroups.tcpping.initial_hosts:localhost[7600],localhost[7601]}"

                           port_range="3"

                           num_initial_members="3"/>

       

                  <MERGE2 max_interval="100000" min_interval="20000"/>

             </stack>

       

           <stack name="jbm-data"

                      ............

                    <TCPPING timeout="3000"

                           initial_hosts="${jgroups.tcpping.initial_hosts:localhost[7600],localhost[7601]}"

                           port_range="3"

                           num_initial_members="3"/>

          </stack>

       

       

      2011-02-08 11:43:19,298 INFO  [org.jboss.ha.framework.server.DistributedReplicantManagerImpl.ess-ldn-sybu1-ETSESSTEST2] Dead members: 0 ([])

       

      2011-02-08 11:43:19,298 INFO  [org.jboss.ha.framework.server.DistributedReplicantManagerImpl.ess-ldn-sybu1-ETSESSTEST2] New Members : 1 ([10.140.1.36:1599])

       

      2011-02-08 11:43:19,298 INFO  [org.jboss.ha.framework.server.DistributedReplicantManagerImpl.ess-ldn-sybu1-ETSESSTEST2] All Members : 9 ([10.140.1.27:1599, 10.140.1.28:1599, 10.140.1.29:1599, 10.140.1.30:1599, 10.140.1.31:1599, 10.140.1.33:1599, 10.140.1.34:1599, 10.140.1.35:1599, 10.140.1.36:1599])

       

      2011-02-08 11:43:19,299 INFO  [org.jboss.cache.RPCManagerImpl] Received new cluster view: [10.140.1.27:48610|8] [10.140.1.27:48610, 10.140.1.28:48610, 10.140.1.29:48610, 10.140.1.30:48610, 10.140.1.31:48610, 10.140.1.33:48610, 10.140.1.34:48610, 10.140.1.35:48610, 10.140.1.36:48610]

       

      2011-02-08 11:43:24,263 ERROR [org.jgroups.blocks.ConnectionTable] failed sending data to 10.140.1.27:48610: java.net.SocketException: Socket closed

       

      2011-02-08 11:43:24,264 ERROR [org.jgroups.blocks.ConnectionTable] failed sending data to 10.140.1.28:48610: java.net.SocketException: Socket closed

       

      2011-02-08 11:43:24,309 ERROR [org.jgroups.blocks.ConnectionTable] failed sending data to 10.140.1.29:48610: java.net.SocketException: Socket closed

       

      2011-02-08 11:43:24,352 ERROR [org.jgroups.blocks.ConnectionTable] failed sending data to 10.140.1.30:48610: java.net.SocketException: Socket closed

       

      2011-02-08 11:43:24,357 ERROR [org.jgroups.blocks.ConnectionTable] failed sending data to 10.140.1.31:48610: java.net.SocketException: Socket closed

       

      2011-02-08 11:43:24,368 ERROR [org.jgroups.blocks.ConnectionTable] failed sending data to 10.140.1.33:48610: java.net.SocketException: Socket closed

       

      2011-02-08 11:43:24,371 ERROR [org.jgroups.blocks.ConnectionTable] failed sending data to 10.140.1.34:48610: java.net.SocketException: Socket closed

       

      2011-02-08 11:43:24,376 ERROR [org.jgroups.blocks.ConnectionTable] failed sending data to 10.140.1.35:48610: java.net.SocketException: Socket closed

       

      2011-02-08 11:43:24,380 ERROR [org.jgroups.blocks.ConnectionTable] failed sending data to 10.140.1.36:48610: java.net.SocketException: Socket closed

       

      2011-02-08 11:43:30,407 INFO  [org.jboss.messaging.core.impl.postoffice.GroupMember] org.jboss.messaging.core.impl.postoffice.GroupMember$ControlMembershipListener@6259e4 got new view [10.140.1.27:48610|8] [10.140.1.27:48610, 10.140.1.28:48610, 10.140.1.29:48610, 10.140.1.30:48610, 10.140.1.31:48610, 10.140.1.33:48610, 10.140.1.34:48610, 10.140.1.35:48610, 10.140.1.36:48610], old view is [10.140.1.27:48610|7] [10.140.1.27:48610, 10.140.1.28:48610, 10.140.1.29:48610, 10.140.1.30:48610, 10.140.1.31:48610, 10.140.1.33:48610, 10.140.1.34:48610, 10.140.1.35:48610]

       

      2011-02-08 11:43:30,407 INFO  [org.jboss.messaging.core.impl.postoffice.GroupMember] I am (10.140.1.27:48610)

       

      2011-02-08 11:43:30,407 INFO  [org.jboss.messaging.core.impl.postoffice.GroupMember] New Members : 1 ([10.140.1.36:48610])

       

      2011-02-08 11:43:30,407 INFO  [org.jboss.messaging.core.impl.postoffice.GroupMember] All Members : 9 ([10.140.1.27:48610, 10.140.1.28:48610, 10.140.1.29:48610, 10.140.1.30:48610, 10.140.1.31:48610, 10.140.1.33:48610, 10.140.1.34:48610, 10.140.1.35:48610, 10.140.1.36:48610])