We are having issues in our live systems because on occassion when downstream consumers are taken off line for maintnenance for any lengthy period of time applications that still post to the queues hang after a number of transactions are posted to the queue.

This seems to be directly related to the memory configuration of MQ, the larger the memory the more messages that can be put in the queue before the situation occurs, but it does always occur.

This issue has existed through many releases of MQ from previous testing.

When MQ reaches it's memory limit, why does it hang instead of rejecting the publish request (we're using producer.send on a persistant queue with no message expiry).

If the client at least could be told the queue is full then we'd have a chance to report the issue and stop the publishers.

Does this issue only occur with the journals. We are moving over to database persistance soon as we hope this removes the issue, will it?