MQSeries.net :: View topic

scott9 · Posted: Thu Dec 18, 2008 12:53 am Post subject:

O/S: AIX 5.1 << old...We know
MQ: Version 5.3 CSD10 << Yeah, we know about this too

Sorry to post about problems in obsolete versions, but my customer is not in a position to upgrade yet. Also, I was not involved in the initial troubleshooting, so I did not see the environment and I have no trace data.

I'm investigating an issue related to a long MQ startup. It normally takes 1 minute to start.
1. The strmqm cmd hung for 2 hrs.
2. They issue endmqm after 2 hrs, but the status remained 'Starting'.
3. They eventually killed the processes and reissued strmqm. It took 1.5 hrs to start on the second attempt.

Only significant FDC is related to busy semaphore for runmqsc. There was also an FDC for busy semaphore on rcrmqobj, but this happened after they killed processes and reissued strmqm.

One thing to note is that there was another qmgr recycled on the system and went into Running status without any issues (seperate filesystem).

My questions:
1. I can't discount the FDC about rcrmqobj...could this just be MQ taking a long time to recover damaged objects? ...and is it possible they damaged the objects by killing the processes on the first attempt?

...if "Yes" to #1, then I still don't have a good reason for the initial delay.

2. Oracle and Message broker are used in this environment and MQ is the XA Coordinator. The endmqm cmd only took 1 minute to complete. If there were any XA rollbacks, I would expect them to occur at shutdown. Is it possible that there were XA rollbacks being processed upon startup?

3. The channel between the qmgr and configuration manager started just prior to the shutdown. Is it possible the config mgr gave the qmgr a batch of work prior to shutdown to be processed at next startup?

4. Any other thoughts on what could keep MQ so busy at startup?