| Author |
Message
|
| matth |
 Posted: Tue Mar 30, 2021 11:46 am Post subject: MQCMIT too slow |
 |
|
Apprentice
Joined: 13 Feb 2007
Posts: 44
Location: Brazil
|
Hi:
I have trouble with an application with Java MQI MQCMIT. An simple call (Connect->Open->Put->Commit), sometimes (3-7 messages in a world of 40000 messages by day).
An Analisys from AppDynamics shows problem:
| Code: |
Servlet - dispatcherServlet:doService: 925(execution time: 0 ms of 6517 ms total)
Servlet - dispatcherServlet:doDispatch: 991(execution time: 0 ms of 6517 ms total)
org.springframework.web.servlet.mvc.method.annotation.RequestMappingHandlerAdapter:handleInternal: 800(execution time: 0 ms of 6517 ms total)
org.springframework.web.servlet.mvc.method.annotation.RequestMappingHandlerAdapter:invokeHandlerMethod: 894(execution time: 0 ms of 6517 ms total)
Spring Bean - boletoEdgeController:insereTitulo: 61(execution time: 0 ms of 6517 ms total)
Proxy For Spring Bean - boletoService:insereBoleto: Unknown (execution time: 0 ms of 6517 ms total)
Spring Bean - org.springframework.transaction.interceptor.TransactionInterceptor#0:invoke: 98(execution time: 0 ms of 6517 ms total)
com.ibm.mq.jms.MQSession:commit: 294(execution time: 0 ms of 6517 ms total)
com.ibm.mq.jmqi.monitoring.JmqiInterceptAdapter:MQCMIT: 247(execution time: 0 ms of 6517 ms total)
com.ibm.mq.jmqi.local.LocalMQ:MQCMIT: 2528(execution time: 0 ms of 6517 ms total)
com.ibm.mq.jmqi.local.internal.base.Native:MQCMIT: Unknown (execution time: 6517 ms of 6517 ms total)
|
This same Application puts an message in another queue, without performance problems. Both queues are DEFPERSIST(NO).
I have compared two QLOCAL and found no differences with them, except IPPROC property, like this:
Q1: IPPROCS (20)
Q2: IPPROCS (0)
What I can do/try to identify this gap and reduce this delay?
Att, |
|
| Back to top |
|
 |
| bruce2359 |
| Posted: Tue Mar 30, 2021 11:53 am Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
Usual questions:
- Is this a new app?
- If it's not a new app, what has changed?
- New changed o/s?
- New or changed anything?
- Does the Unit-of-Work include a data base update?
- Any errors posted to MQ error logs for this qmgr?
I noticed that you posted this to the Telemetry forum. How is this related to telemetry?
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| gbaddeley |
| Posted: Tue Mar 30, 2021 1:36 pm Post subject: |
|
|
Jedi Knight
Joined: 25 Mar 2003
Posts: 2538
Location: Melbourne, Australia
|
The queue manager's recovery logs could be under stress. Are you using linear logs or circular logs? How quickly does a log file turn over?
Are the IPPROCS also using MQCMIT? How long is their unit of work?
_________________
Glenn |
|
| Back to top |
|
|
| bruce2359 |
| Posted: Tue Mar 30, 2021 3:34 pm Post subject: |
|
|
Poobah
Joined: 05 Jan 2008
Posts: 9469
Location: US: west coast, almost. Otherwise, enroute.
|
Look at MQ.ini file for this qmgr. Post all log values.
_________________
I like deadlines. I like to wave as they pass by.
ב''ה
Lex Orandi, Lex Credendi, Lex Vivendi. As we Worship, So we Believe, So we Live. |
|
| Back to top |
|
|
| matth |
| Posted: Wed Mar 31, 2021 8:54 am Post subject: |
|
|
Apprentice
Joined: 13 Feb 2007
Posts: 44
Location: Brazil
|
Folks, tks in all for help. I´ll reply all questions in this post:
Log properties:
| Code: |
Log:
LogPrimaryFiles=27
LogSecondaryFiles=13
LogFilePages=65535
LogType=CIRCULAR
LogBufferPages=512
LogPath=/var/mqm/log/QMNAME/
LogWriteIntegrity=TripleWrite
|
- Is this a new app? If it's not a new app, what has changed?
R: No. It has not been changed (unit-of-work) al least 2 years ago.
- New changed o/s?
R: O/S patching is frequent. weekly probably, but I don´t have acess about it.
- New or changed anything?
R: N/A
- Does the Unit-of-Work include a data base update?
R: No.
- Any errors posted to MQ error logs for this qmgr?
R: Requesting logs yet.
-Are the IPPROCS also using MQCMIT?
R: All threads are the same as problematic case. Repeat. This don´t ocour for all messages.
-How long is their unit of work?
R: Usually, it takes 0.050ms from begin to Commit state.
In add, App Developer ask me if two parallel threads may cause this like this:
T1: Open
T1: Put
T2: Open
T2: Put
T1: Commit
T2: Commit
This may cause some lock on Queue which causes this delay to happen?
Att, |
|
| Back to top |
|
|
| hughson |
| Posted: Thu Apr 01, 2021 12:26 am Post subject: |
|
|
Padawan
Joined: 09 May 2013
Posts: 1959
Location: Bay of Plenty, New Zealand
|
It sounds like you are telling us that you have lots of contention on this queue, that is, there are lots of connections using this queue.
You may have read about a recently-ish (V9.0.5) added feature in MQ called Implicit Syncpoint. It was added because the contention on a queue is longer if a transaction is not used since the queue manager has to lock the queue for longer.
There is a good blog post about this (that was saved from a fiery death when IBM killed off previous blogging platforms) and it can now be read here.
So my question to you is, could some of these interactions with this queue be outside of a transaction? Also what version of MQ are you using since this feature is ON by default so you should be getting the benefit from it if you are at or above that level.
Cheers,
Morag
_________________
Morag Hughson @MoragHughson
IBM MQ Technical Education Specialist
Get your IBM MQ training here!
MQGem Software |
|
| Back to top |
|
|
| exerk |
| Posted: Thu Apr 01, 2021 1:11 am Post subject: |
|
|
Jedi Council
Joined: 02 Nov 2006
Posts: 6339
|
Moved to a more appropriate forum.
_________________
It's puzzling, I don't think I've ever seen anything quite like this before...and it's hard to soar like an eagle when you're surrounded by turkeys. |
|
| Back to top |
|
|
| matth |
| Posted: Thu Apr 01, 2021 10:42 am Post subject: |
|
|
Apprentice
Joined: 13 Feb 2007
Posts: 44
Location: Brazil
|
Tks Morag Hughson:
It looks an good approach on problem. I´ll check right version of WMQ.
Att, |
|
| Back to top |
|
|
|
|
