OpenMQ Unack'ed messages causing Java Heap overflow in STOMP

Question

Hello All,
I am using OpenMQ and STOMP within .Net using C#.
Using OpenMQ's imqcmd.exe command I can see messages being produced for a given consumed topic. I have AutoAcnowledge set and have tried sending messages but still see the UnAck message count grow and grow until my STOMP application runs out of heap space and then the world quickly comes to a grinding halt.
I have tried sending the messages as Non Persistent, have tried limiting the Topic message store size to 1000 messages, and have tried a session.commit after each send but cannot seem to figure out 1) why these messages are UnAck'd in the first place and 2) How to keep the STOMP Java app from getting an overflow and ceasing to function.
Thanks for any help that any of you may be able to offer,
Ted Squilanti

Answer 1

We've logged this as Glassfish issue 11321.
The bug is related to the way in which Glassfish starts the broker lazily when you first try to connect to it.
There is a workaround, which involves changing the lazy-init attribute of the jms-service element in domain.xml to false:
      <jms-service default-jms-host="default_JMS_host" type="EMBEDDED">
        <jms-host host="localhost" name="default_JMS_host" lazy-init="false" />
      </jms-service>This disables the mechanism which starts the embedded broker automatically when a client first tries to connect. Instead, the broker will be started when Glassfish tries to lookup a JMS resource. The simplest way to achieve this is to call "asadmin jms-ping" .
Nigel