Author |
Message
|
ColinRenouf |
Posted: Fri Sep 12, 2003 11:10 pm Post subject: Capacity Units and setmqcap |
|
|
Newbie
Joined: 12 Sep 2003 Posts: 6 Location: London, England
|
Can someone please tell me if capacity units and the use of setmqcap actually affect the behaviour of MQ 5.3 in the way it uses processors? _________________ Colin Renouf |
|
Back to top |
|
 |
PeterPotkay |
Posted: Sat Sep 13, 2003 2:44 pm Post subject: |
|
|
 Poobah
Joined: 15 May 2001 Posts: 7722
|
If you don't set it correctly (high enough), MQ wont run properly (based on Solaris expirience).
I have never tried setting the number higher, but it definitly does not work if it is lower. _________________ Peter Potkay
Keep Calm and MQ On |
|
Back to top |
|
 |
ColinRenouf |
Posted: Sat Sep 13, 2003 3:56 pm Post subject: |
|
|
Newbie
Joined: 12 Sep 2003 Posts: 6 Location: London, England
|
Peter,
Does this sound familiar? A live 12-way AIX box on which this was never set where messages replied to the box build up and then "pulse" into it periodically in large batches, and CPU 0 being considerably busier than the others. The system test box did have setmqcap run and doesn't exhibit this behaviour, but as the live box is key to the business I need a good reason to bounce MQ on it!
Cheers and thanks
Colin _________________ Colin Renouf |
|
Back to top |
|
 |
PeterPotkay |
Posted: Sat Sep 13, 2003 4:22 pm Post subject: |
|
|
 Poobah
Joined: 15 May 2001 Posts: 7722
|
I have never seen this behaviour. But if setmqcap is not correct, I would not bother trying to debug ANY mq problems until the setmqcap is correct.
Issue dspmqcap to see what it currently set at and what MQ thinks is the number of proccessors on the machine (and what # you should use for setmacap). _________________ Peter Potkay
Keep Calm and MQ On |
|
Back to top |
|
 |
ColinRenouf |
Posted: Wed Sep 17, 2003 10:52 pm Post subject: |
|
|
Newbie
Joined: 12 Sep 2003 Posts: 6 Location: London, England
|
Peter,
IBM say there is definitely nothing in the code that does anything in response to the capacity units, but doing a dump of the code on NT (and you can see it is common code) there is a call to SetProcessAffinityMask in the code that handles the threading.
Thanks for your help. If IBM say it doesn't do anything who am I to argue - even if the code disagrees?
Note that the NT DUMPBIN /imports call on the executables for setmqcap and the actual queue managers is VERY effective at seeing what the code does. There are similar calls on AIX which I have used in the past but I don't have the authority to run them on our AIX machine. _________________ Colin Renouf |
|
Back to top |
|
 |
|