MQSeries.net :: View topic

PeterPotkay · Posted: Mon Aug 05, 2002 7:26 pm Post subject:

One way this can happen is if the application that gets triggered never opens the queue. It blew up before the MQOPEN for whatever reason.
Now we have 1 message on the queue and no MQCLOSE happening to force another trigger. As more message land, no more triggers, since the queue depth is not going from 0 to 1. The trigger interval save us in this spot, assuming the reason for the app blowing up before the MQOPEN was temprorary.

Or,
On the mainframe, say your app abends after retrieving the message. Somehow a few bad messages got on your queue. Your poisened message logic removes them after a couple of GETS cause the backout count exceeded your threshhold. Each time the message gets rolled back, the app is retriggered because of the MQCLOSE. Eventually you clear out the 2 or 3 bad messages to the exception queue, but in the meantime you abended so many times that your TCODE gets brought down. Now the trigger monitor can't start the app, but more (good) message keep landing on the queue. Again we have a situation where the queue is closed , the depth is going from 0+n to 0+n+1, and no trigger messages are being generated. The TCODE is restarted. If the trigger interval passes, the app will now automaticaly start up again.

The init queue can be full, in which case the trigger message go to the DLQ. Again as more message land on that closed triggered on first queue, no more trigger messages get generated, even as the init queue's depth came down.

There were a couple more situations that I came across a while back, but they escape me now. By the way, the trigger interval only applies to Trigger First queues, and since it is set at the QM level, you have to pick a value that keeps all your applications happy. (We chose 5 minutes (300000)).
_________________
Peter Potkay
Keep Calm and MQ On