MQSeries.net :: View topic

bigdavem · Posted: Wed Oct 09, 2002 4:20 am Post subject:

I have 10 NT workstations connecting as v5 MQ clients to a v5.1 queue manager, also on NT. These workstations all compete for messages on a single queue.

I start with nothing on the queue. I then begin sending 800 messages of varying sizes (from 20Kb up to 2Mb). Each message contains a unique ID. Of the 800 messages, about half of the first 20 (seemingly selected at random) are processed multiple times (usually twice, sometimes three times). I can track this because the applications log the unique ID and also the MsgId.

The applications all browse the queue on a two second wait, then break out and perform some other processing before browsing again. The LOCK and BROWSE_FIRST options are set. When the application succesfully gets a message, it checks the data to make sure it's a message it wants then issues a get, MSG_UNDER_CURSOR and SYNCPOINT. It then does its processing (betwen 5 minutes and 20 hours) and commits the message upon success.

As an example, machine 1 picks up message 1 at 1:17:39am and finishes processing at 1:38:37am. Meanwhile, machine 2 picks up message 1 at 1:30:39am and finishes processing at 2:00:00am.

This doesn't make sense to me. Surely machine 1 has message 1 exclusively at all times? So how could machine 2 get its hands on it? Also, why does it only happen for the first 20 messages out of 800? Each message is definitely only on the queue once ie the sending application is not duplicating messages (as evidenced by each instance of a message having the same MsgId logged).

I'm stumped. Help? I can provide more info if needed, just let me know what!

Thanks,

Dave.