Distributed Destinations and Error Queues

Question

          I have a distributed queue - distqueue which is configured with two queue members
          - srvr1q
          and srvr2q. srvr1q is a physical queue destination that is running on JMSServer1
          on one instance in a cluster. srvr2q is a physical queue destination that is running
          on JMSServer2 on another instance in the cluster. I have two errorqueues - srvr1errq
          and srvr2errq - deployed on JMSServer1 and JMSServer2 respectively. I have configured
          the srvr1q to the following:
          Redelivery Delay Override: 10000
          Redelivery Limit: 3
          Error Destination: srvr1errq
          Also the srvr2q is configured to the following:
          Redelivery Delay Override: 10000
          Redelivery Limit: 3
          Error Destination: srvr2errq
          There is an MDB listening to each one of these local queues. The MDB is programmed
          to throw a null pointer exception (to simulate business logic processing failures)
          My expectations are : the message goes to one of the local queues. Null Pointer
          exception is thrown. After 10 seconds the message is resent to the same queue
          with a similar fate. After the third attempt, the message is moved to the error
          queue of the corresponding local queue.
          What I see is the following:
          When I send a message I do see a Null Pointer Exception, but it occurs only once
          and I dont see the message going to the error queue. Since my Load Balancing policy
          is Round-Robin on the distqueue, the local queues receive the message every other
          time with the same result - Null Pointer Exception thrown only once.
          I also have JMSStores configured so the messages are persistent.
          Can someone explain to me what I am doing wrong?
          Thanks.
          

Answer 1

          "Prakash Muthukrishnan" <[email protected]> wrote:
          >
          >Hi Tom
          > Thanks again for the reply. I do use a tx MDB. I have set the transaction-type
          >to Container in the message-driven stanza in ejb-jar.xml. And also the
          >trans-attribute
          >for that EJB in the container-transaction stanza, I have used Required.
          >Do I need
          >to setup anything else ?
          I don't think so.
          > The transaction descriptor element in the weblogic-enterprise-bean
          >stanza in weblogic-ejb-jar.xml for that MDB is empty.
          Not sure what you mean?
          > Am I missing anything
          >else?
          Container & Required is all that is needed
          If you like, post your ejb descriptor xml file, and I'll take a quick
          look to double-check.
          >-Prakash
          >
          >Tom Barnes <[email protected]> wrote:
          >>Hi Prakash,
          >>
          >>A runtime exception should force a rollback for tx MDBs, and
          >>a recover for non-tx MDBs. If you are seeing different
          >>behavior there may be a bug in WL.
          >>
          >>Since MDBs do not log an exception when the app throws
          >>a Runtime exception, I often recommend wrapping the
          >>application code in a try catch anyhow
          >> try {
          >> } catch (Throwable t) {
          >> log
          >> force rollback
          >> }
          >>I think there are plans to provide a configurable that
          >>optionally does something similar in the next release (no guarantees).
          >>(The problem is that some apps throw a runtime on purpose, and
          >>could quickly flood the log if such runtimes were logged per message.)
          >>
          >>Tom
          >>
          >>Prakash Muthukrishnan wrote:
          >>> Tom
          >>> Thanks for the suggestion. Using the setRollbackOnly method did the
          >>trick. This
          >>> brings up
          >>> a related question: What if a run time exception was thrown during
          >>the processing
          >>> in an MDB?
          >>> Is it not retried and ultimately shoved into the error queue?
          >>> Thanks again.
          >>> -Prakash
          >>>
          >>>
          >>> Tom Barnes <[email protected]> wrote:
          >>>
          >>>>Hi Prakash,
          >>>>
          >>>>At first glance, it doesn't seem like your doing anything wrong.
          >>>>Could you post your release and SP level, as well as your config.xml?
          >>>>
          >>>>How many messages did you send through?
          >>>>
          >>>>Note that for transactional MDBs it is best to call "setRollbackOnly"
          >>>>on the EJB context rather than throwing an NPE. This is because
          >>>>throwing a runtime exception forces the MDB to destroy itself and
          >>>>get re-initialized (pretty high overhead). For non-transactional
          >>>>MDBs, throwing a runtime exception is your only option.
          >>>>
          >>>>Tom, BEA
          >>>>
          >>>>Prakash Muthukrishnan wrote:
          >>>>
          >>>>
          >>>>>I have a distributed queue - distqueue which is configured with two
          >>>>
          >>>>queue members
          >>>>
          >>>>>- srvr1q
          >>>>>and srvr2q. srvr1q is a physical queue destination that is running
          >>>>
          >>>>on JMSServer1
          >>>>
          >>>>>on one instance in a cluster. srvr2q is a physical queue destination
          >>>>
          >>>>that is running
          >>>>
          >>>>>on JMSServer2 on another instance in the cluster. I have two errorqueues
          >>>>
          >>>>- srvr1errq
          >>>>
          >>>>>and srvr2errq - deployed on JMSServer1 and JMSServer2 respectively.
          >>>>
          >>>>I have configured
          >>>>
          >>>>>the srvr1q to the following:
          >>>>>
          >>>>>Redelivery Delay Override: 10000
          >>>>>Redelivery Limit: 3
          >>>>>Error Destination: srvr1errq
          >>>>>
          >>>>>Also the srvr2q is configured to the following:
          >>>>>Redelivery Delay Override: 10000
          >>>>>Redelivery Limit: 3
          >>>>>Error Destination: srvr2errq
          >>>>>
          >>>>>There is an MDB listening to each one of these local queues. The
          >MDB
          >>>>
          >>>>is programmed
          >>>>
          >>>>>to throw a null pointer exception (to simulate business logic processing
          >>>>
          >>>>failures)
          >>>>
          >>>>>My expectations are : the message goes to one of the local queues.
          >>>>
          >>>>Null Pointer
          >>>>
          >>>>>exception is thrown. After 10 seconds the message is resent to the
          >>>>
          >>>>same queue
          >>>>
          >>>>>with a similar fate. After the third attempt, the message is moved
          >>>>
          >>>>to the error
          >>>>
          >>>>>queue of the corresponding local queue.
          >>>>>
          >>>>>What I see is the following:
          >>>>>When I send a message I do see a Null Pointer Exception, but it occurs
          >>>>
          >>>>only once
          >>>>
          >>>>>and I dont see the message going to the error queue. Since my Load
          >>>>
          >>>>Balancing policy
          >>>>
          >>>>>is Round-Robin on the distqueue, the local queues receive the message
          >>>>
          >>>>every other
          >>>>
          >>>>>time with the same result - Null Pointer Exception thrown only once.
          >>>>>
          >>>>>I also have JMSStores configured so the messages are persistent.
          >>>>>
          >>>>>Can someone explain to me what I am doing wrong?
          >>>>>Thanks.
          >>>>
          >>>
          >>
          >