1 Reply Latest reply on Apr 4, 2012 1:05 PM by David Barr

    Persistence not accessible by the whole cluster?

    David Barr Newbie

      Good Morning,


      I need some clues on how to troubleshoot this. I'm only just beginning to examine the issue, so I'll even take hints on how my initial guesses can be refuted for better ideas. Here's what appears to be happening:


      • Two-node JBoss EAP 4.3.3 cluster with queues set and mysql-ds.xml persistence as DefaultDS.
      • Queue sender connects to one node.
      • Queue receiver connects to the other node.
      • Queue receiver can't receive messages.
      • JMX-Console on Node1 shows 50+ messages in the queue. JMX-Console on Node2 shows 0 messages in the same queue.


      Now, if (stop/start receiver) && (receiver reconnects to same host as sender) message passing works fine. The problem only shows up if the sender and receiver are on different hosts in the cluster. The server.log files on both nodes don't show me anything obviously interesting, but I might be missing a relevant message. As far as I can tell, both nodes are connecting to the persistence database without a problem.


      The files deploy/mysql-ds.xml, deploy/jboss-messaging.sar/mysql-persistence-service.xml, and deploy/hajndi-jms-ds.xml all appear reasonable to me. But, I'm not a JBoss guy.


      What should I be looking for in server.log? What else should I be looking for? Can I craft a test within the cluster to identify why they don't both appear to be seeing into the persistence database?




        • 1. Re: Persistence not accessible by the whole cluster?
          David Barr Newbie

          [Troubleshoot access to the MySQL persistence database]


          I did just check and find that the persistence user given in mysql-ds.xml only had access privileges from one of the two hosts in the cluster. I gave that user privileges to connect from the other host, and found messages in both server.logs that tell me both nodes are connecting to the MySQL database.


          However, I'm still not seeing matching queue counts for the same queue on both nodes.