2 Replies Latest reply on Nov 13, 2006 8:40 AM by Marcio Cossari

    Transaction locking issue on JBoss 3.2.6 under heavy load in

    Robert Costanzo Newbie


      I am running JBoss 3.2.6 in a 5 instance cluster. I am using the container configuration cmp2.x jdbc2 pm with cache invalidation. I am NOT doing any HTTP session replication. When my servers are under heavy load, I experience the following issue:

      One instance will start logging errors like:

      WARN [org.jboss.tm.TransactionImpl] Transaction TransactionImpl:XidImpl [FormatId=257, GlobalId=mateso-webapp3//16199095, BranchQual=] timed out. status=STATUS_ACTIVE
      WARN [org.jboss.tm.TransactionImpl] Lock contention, tx=TransactionImpl:XidImpl [FormatId=257, GlobalId=mateso-webapp3//16199095, BranchQual=]

      and become locked up. Subsequently, all of the other 4 instances in my cluster become locked up. I have to kill the instance that the locking issue originated on for all of the other instances to start working properly again.

      I have checked our Oracle database and there are no waiting queries when this happens, which points to it not being a database locking issue.

      What other possibilities are there which would cause all members of the cluster to lock? I had thought that cache invalidation was asynchronous, but if it was synchronous then that could explain it. Also, is there a way to track down what the transaction warnings above are locking on?