|
RSS Feed - WebSphere MQ Support
|
RSS Feed - Message Broker Support
|
 |
|
High swap memory utilization by EG's |
« View previous topic :: View next topic » |
Author |
Message
|
bloomy |
Posted: Thu Apr 22, 2010 10:18 am Post subject: High swap memory utilization by EG's |
|
|
 Acolyte
Joined: 11 Feb 2009 Posts: 61 Location: Bloomington, IL USA
|
Hello All,
We are running WMB 6104 on AIX 5.3. we are seeing issues with a particular framework message flows consuming very high memory(swap) utilization for the EG's that have these flows. These flows are very large with 150 nodes per each flow and 15 to 20 instances of each flow running in every EG and almost 10 of such flows. We tried to isolate the flows to their own EG's but they are consuming equally high memory even in a scenario where we have one such flow per EG.
I know it is a vague question to ask for the reason for the high memory, because without seeing the code no one can say why they are consuming so. But my reason for posting here is to know if anyone else has seen similar issues in their envs and what might a few possibilities for high memory. May be if possible a few references from IBM docs or anywhere else which can point us to similar situations.
FYI: we opened a PMR and worked for 3 months with IBM hursely folks and they have asked us to set an env variable MQSI_FREE_MASTER_PARSER, initially for a few days we were comfortable with the memory, but looks like the problem is back after the appication area made some code changes.
One such change to use File input node sunsetting the MBFE.
Thanks in advance. |
|
Back to top |
|
 |
mqjeff |
Posted: Thu Apr 22, 2010 11:05 am Post subject: |
|
|
Grand Master
Joined: 25 Jun 2008 Posts: 17447
|
The usual causes of high memory utilization are excessive tree copying and excessive use of field[i] paths instead of reference variables.
A good trick to avoid excessive tree copying is to start your flow by putting the input message into a properly formed parser tree in Environment, and then do all your manipulation and transformation against that, and never call CopyEntireMessage() or CopyMessageHeaders(). |
|
Back to top |
|
 |
WMBDEV1 |
Posted: Thu Apr 22, 2010 11:09 am Post subject: |
|
|
Sentinel
Joined: 05 Mar 2009 Posts: 888 Location: UK
|
I'm not sure what I can offer you that Hursley havent already but I have worked on flows that have input messages of 200MB (compressed to 33MB) which I then transformed and output to a message of similar size. Just out of interest, how big are your input messages and why do you have so many nodes in your flow?
Also, how much memory are the flows using? FWIW, My flow (1 instance) ended up using about 700MB. I could have reduced this down by reducing the nodes used in some developer purity activity but it wasnt necessary in the end.
Have you tried to isolate cerain areas of the flow that are memory intensive or does each node increase the size by a small amount but due to the volume of them this becomes significant? |
|
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
|
|
|
|