6 Replies Latest reply on Jan 29, 2004 9:31 PM by Tim McCune

    Delayed joining of partition

    Michael Daleiden Newbie

       

      "michael.daleiden" wrote:
      I have a JBoss cluster set up between two nodes (Win2K and HP/UX). When I start the first node, everything goes smoothly. When I then go to start the second node, things get strange. It appears from the logs on the second node that the second node does not immediately join the partition (the log reports only one partition member):

      2003-08-06 11:14:05,838 INFO [org.jboss.ha.framework.server.ClusterPartition] Starting channel
      2003-08-06 11:14:05,839 INFO [org.jboss.ha.framework.interfaces.HAPartition.DefaultPartition] Number of cluster members: 1
      2003-08-06 11:14:05,840 INFO [org.jboss.ha.framework.interfaces.HAPartition.DefaultPartition] Other members: 0

      Later on in the log (after the server startup is complete and just before the farming service starts polling for new deployments), it reports that the cluster view has changed to include the other node:

      2003-08-06 11:15:46,562 INFO [org.jboss.ha.framework.interfaces.HAPartition.DefaultPartition] New cluster view: 1 ([mdaleiden2:1705, codvap02:51245] delta: 1)
      2003-08-06 11:15:46,568 INFO [DefaultPartition:ReplicantManager] Merging partitions...
      2003-08-06 11:15:46,568 INFO [DefaultPartition:ReplicantManager] Dead members:0

      Configuration:
      Win2K node: Win2K SP2, JDK 1.4.1_01, JBoss 3.2.2RC1
      HP/UX node: HP/UX 11.0, JDK 1.3.1_08, JBoss 3.2.2RC1

      Nodes are on separate subnets, but multicast is enabled across the routers.

      Any ideas as to why this delay occurs? I really need to get the clustering issues worked out ASAP, as we are trying to get prepared for a pilot launch of our application.