Compensating for JMS on one machine

Question

          So i'm still coming up to speed on clustering with Weblogic,
          and I have a few questions.
          1) how (if at all) are people compensating for the reliability
          issue raised by running JMS on only one machine in the cluster.
          2) is there a way to make prevent the one machine running JMS
          from performing any tasks normal other than JMS? ie: prevent
          it from accepting EJB requests, JSP request, etc. & have it
          focus only on JMS (arguably to reduce the chances that some
          errant code unrelated to JMS might bring the system down
          thus choking the rest of the cluster)
          3) Does weblogic have plans to make JMS a clustered service?
          thanks,
          jason
          

Answer 1

Joel,
          Just to clarify... So, by suffixing qnames with unique ids you are
          effectively trying to mimic load-balancing by the clients, since Nodes ids
          are going to be known to the client apriori. Is it correct?
          Thanks
          Daniel Ilkanayev
          [email protected]
          Joel Nylund <[email protected]> wrote in message
          news:[email protected]...
          > Hey Jason,
          >
          > some people from weblogic in the past have recomended putting all your
          > q's in one of the boxes on the cluster. I dont think this is a good idea
          > since you loose one of the benefits of clustering when you do this
          > (fault tolerance).
          >
          > One idea I have used is to put the same queues in each server. If one
          > server goes down, the messages will be left on the queue (if they are
          > persistent). If you couldnt get the server back up in a reasonable
          > amount of time you could always point another server or process at those
          > q's and consume them. One problem I have had with this approach is that
          > even though weblogic claims jms is not a clustered service, they have
          > compiled their stubs with replication on (at least in 4.5.1 sp7). This
          > causes problems if you use the same jndi name for q's. You will see the
          > conflit handler go nuts if you have 2 servers with the same q name. We
          > have worked around the problem by suffixing our qnames with uniqe ids, I
          > think this will work (going to test tonight).
          >
          > If anyone else has any better ideas for using jms in a enterprise system
          > with fault tolerance and load balancing, please let me know.
          >
          > -Joel
          >
          >
          >
          > Jason Mellein wrote:
          >
          > > So i'm still coming up to speed on clustering with Weblogic,
          > > and I have a few questions.
          > >
          > > 1) how (if at all) are people compensating for the reliability
          > > issue raised by running JMS on only one machine in the cluster.
          > >
          > > 2) is there a way to make prevent the one machine running JMS
          > > from performing any tasks normal other than JMS? ie: prevent
          > > it from accepting EJB requests, JSP request, etc. & have it
          > > focus only on JMS (arguably to reduce the chances that some
          > > errant code unrelated to JMS might bring the system down
          > > thus choking the rest of the cluster)
          > >
          > > 3) Does weblogic have plans to make JMS a clustered service?
          > >
          > > thanks,
          > >
          > > jason
          >