2 Replies Latest reply on Nov 11, 2014 3:12 PM by Mark Lybarger

    separate persistence for queue vs dlq

    Mark Lybarger Apprentice

      can i use a separate persistence adapter for my "production" queues .vs. the dlq? i'd like to use pgsql for the dlq and kahadb for the producion queue.

        • 1. Re: separate persistence for queue vs dlq
          Jochen Cordes Novice

          Hi Mark,

           

          Unfortunately mixing different kind of persistence stores per Broker is not possible at the moment. For kahadb you can at least split the different destinations into separate journals (aka "Filtered KahaDB": https://access.redhat.com/documentation/en-US/Red_Hat_JBoss_A-MQ/6.1/html/Configuring_Broker_Persistence/files/FuseMBMul…).

           

          Is the reason you want to store the DLQ into PostgrSQL that you need to have a central, high available storage in case there are issues with processing messages?

           

          Kind regards,

           

          Jochen

          • 2. Re: separate persistence for queue vs dlq
            Mark Lybarger Apprentice

            Thanks so much for your response!

             

            It would be a very nice feature to be able to use different stores, even per queue, or grouping of queueus.

             

            It was mentioned that one possible option would be to create a Policy that can be applied instead of the sharedDeadLetterStrategy.  This would allow us to persist the messages before posting to dlq.

             

            The desire to have separate stores is to avoid a problem where a queue can affect all the queues on the broker by way of dlq.  Say I have two queues, A and B.  The processors of messages on A are unable to commit the work due to some unknown reason.  That eventually sends the messages to dlq.  dlq can fill up disk space and memory usage causing queue B to become unusable.  No messages can be "lost" in the system.  Using a separate store allows the system's exception processing to have much less impact on the normal (production) throughput.