| |
|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
  |
|
| WebSphere AS Sessions hammering a Qmgr |
« View previous topic :: View next topic » |
| Author |
Message
|
| januckolls |
 Posted: Mon Jul 17, 2006 9:36 am Post subject: WebSphere AS Sessions hammering a Qmgr |
 |
|
Newbie
Joined: 15 May 2003
Posts: 9
|
While running the MA0W API Crossing Exit against a Qmgr that was supplying messages to a WebSphere AS Connection Pool/JMS Listener/Message Driven Bean setup using client connections I noticed threads running on the MQ server that had an MQGET loop with a WaitInterval of 100 ms. Some were getting messages under syncpoint (with a path thru MQCMIT), others were not under syncpoint (no MQCMIT call). I am making the assumption that these were associated with Sessions defined under AS. Obviously, the Qmgr is getting hammered (server running at 80% CPU).
Not being a WebSphere AS whiz, I went searching for where you configure the WaitInterval in the Configuration Management panels and found nothing even remotely resembling that kind of control. Am I blind, or are we supposed to control the load on the MQ server through things like Reap Time, Unused Timeout, and Aged Timeout? That doesn’t seem like a very efficient way to control the number of unproductive MQGET’s being issued against the Qmgr.
I would really like to get at that MQGET WaitInterval!!
Cheers…
Jim Nuckolls |
|
| Back to top |
|
 |
| jefflowrey |
| Posted: Mon Jul 17, 2006 9:41 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
Ask your developers to please not use JMS selectors.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| januckolls |
| Posted: Mon Jul 17, 2006 11:29 am Post subject: |
|
|
Newbie
Joined: 15 May 2003
Posts: 9
|
I would expect Get with Browse if JMS Selectors were in use that queried anything in the message (including the MQRH2 header) and these requests do not specify a browse operation in the GetMessageOptions. I will go back and check the MQMD structure prior to the get to see if a provider-specific MsgID or CorrelID was specified. I don't mind provider-specific fields but I do mind the 100 ms WaitInterval.
Cheers...
Jim |
|
| Back to top |
|
|
| januckolls |
| Posted: Fri Jul 21, 2006 11:53 am Post subject: |
|
|
Newbie
Joined: 15 May 2003
Posts: 9
|
For those that are interested in who the culprit was -
The listeners were specified as Non-ASF with the Non-ASF Timeout specified as 100 milliseconds. The timeout value is supplied as the WaitInterval on the MQGET. That interval was increased to 5,000 milliseconds which drastically reduced CPU utilization on the WebSphere AS machine as well as the Qmgr server machine. On the Qmgr side CPU utilization went from 52% to 21%  The WebSphere AS documentation suggests setting this to a value that is slightly smaller than the transaction timeout value. For example, if transaction timeout is set to 120 seconds, then set Non-ASF Timeout to 110 seconds. Even though we tried the 110 second timeout value, some other setting (don't know which one yet) did not allow the value to increase past the 5 second value.  |
|
| Back to top |
|
|
|
|
|
| |
|
Page 1 of 1 |
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
|
|
|
