Uploaded image for project: 'JBoss A-MQ'
  1. JBoss A-MQ
  2. ENTMQ-2099

Producer Flow Control Does Not Kick In for Small Message Sizes

    XMLWordPrintable

Details

    • Bug
    • Resolution: Done
    • Major
    • None
    • JBoss A-MQ 6.2.1, JBoss A-MQ 6.3
    • broker
    • Hide

      Install and start a vanilla instance of JBoss A-MQ.
      Start an internal producer with default arguments OR start an external producer with very small payload size (I used a 15 byte random string):

      JBossA-MQ:karaf@root> activemq:producer --user admin --password admin

      In another client session, watch dstat. Run the producer several times and the stats hang - message counts and enqueue counts increase, but memory usage stays frozen. Looking into JMX, we see caching became disabled before PFC could kick in. I tried this with cursorMemoryHighWaterMark values of 50%, 70%, and 150%.

      Show
      Install and start a vanilla instance of JBoss A-MQ. Start an internal producer with default arguments OR start an external producer with very small payload size (I used a 15 byte random string): JBossA-MQ:karaf@root> activemq:producer --user admin --password admin In another client session, watch dstat. Run the producer several times and the stats hang - message counts and enqueue counts increase, but memory usage stays frozen. Looking into JMX, we see caching became disabled before PFC could kick in. I tried this with cursorMemoryHighWaterMark values of 50%, 70%, and 150%.

    Description

      When using the built-in producer to test A-MQ functionality (for example, PFC), memory usage statistics for the given queue become frozen and PFC never kicks in. This appears to be due to caching becoming disabled and seems to happen with a small message size. The default behavior of the activemq:producer seems to be to sent a text payload of around 15 bytes. If a larger message size is specified, PFC seems to kick in fine. I later verified the same behavior with an external producer and small payload.

      JBossA-MQ:karaf@root> activemq:producer --user admin --password admin
JBossA-MQ:karaf@root> dstat
Name                                                Queue Size  Producer #  Consumer #   Enqueue #   Dequeue #   Forward #    Memory %
ActiveMQ.Advisory.Connection                                 0           0           0           2           0           0           0
ActiveMQ.Advisory.MasterBroker                               0           0           0           1           0           0           0
ActiveMQ.Advisory.Producer.Queue.TEST                        0           0           0           2           0           0           0
ActiveMQ.Advisory.Queue                                      0           0           0           1           0           0           0
TEST                                                      1000           0           0        1000           0           0          69
JBossA-MQ:karaf@root> activemq:producer --user admin --password admin
JBossA-MQ:karaf@root> dstat
Name                                                Queue Size  Producer #  Consumer #   Enqueue #   Dequeue #   Forward #    Memory %
ActiveMQ.Advisory.Connection                                 0           0           0           4           0           0           0
ActiveMQ.Advisory.MasterBroker                               0           0           0           1           0           0           0
ActiveMQ.Advisory.Producer.Queue.TEST                        0           0           0           4           0           0           0
ActiveMQ.Advisory.Queue                                      0           0           0           1           0           0           0
TEST                                                      2000           0           0        2000           0           0          69

      Attachments

        Issue Links

          is related to

          Bug - A problem that impairs or prevents the functions of the product. ENTMQ-2103 [A-MQ6, producer flow control, destination memory usage] destination memory usage incorrect after sendtimeout due to PFC.

          • Major - To be worked after higher priority work items are addressed. When the severity is high and the effort to change it is low to moderate, this term is chosen. Issues in this category likely have an existing workaround but implementation or execution may be non-trivial.
          • Closed

          Activity

            People

              gtully@redhat.com Gary Tully
              rhn-support-dhawkins Duane Hawkins
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: