Infrastructure at your Service

Middleware Team

Documentum story – Add the Java urandom inside the script to improve the performance of all WLST sessions

WebLogicServer takes a long time (e.g. over 15 minutes) to startup. Can the performance be improved?
Using /dev/urandom option in the Weblogic Java Virtual Machine parameters from startup script, can be the solution.
In fact, /dev/random is a blocking device and during times of low entropy (when there is not enough random bits left in it), /dev/random will block any processes reading from it (and hence hang) until more random bits have been generated.
So, if you are running some custom WLST scripts to startup your Weblogic infrastructure, you have to know that there are a few things you could try to improve performance.

This change allows to improve the performances of your WLST sessions.. There is no functional impact to modify the script.

1- Login to the Weblogic Server.
2- Use the Unix (installation owner) account:
3- Go to the following directory using:

cd $ORACLE_HOME/oracle_common/common/bin

4- Edit the script with vi editor
5- Add the following option in JVM_ARGS


6- Please note that if you specifically want to reduce the load time in start time script, you can do It by providing :

java weblogic.WLST -skipWLSModuleScanning <script>.py 

It can be an advantage in some cases.

Others ways to increase performance of your WLS:

1- This blog deals with urandom parameter especially.
Be aware that you have another ways  to tune well your Weblogic Server(s).
Just in order to give you an overview of all possible changes :
Set Java parameters in starting Weblogic Server script or directly in JAVA_OPTS:
• urandom (see above)
• java heap size :
For higher performance throughput, set (if possible) the minimum java heap size equal to the maximum heap size. For example:

"$JAVA_HOME/bin/java" -server –Xms512m –Xmx512m -classpath $CLASSPATH ... 

2- Think about the number of Work Managers you’re really needed :
A default Work Manager might not be adapted to all application’s needs.
In this case, you’ll have to create a custom Work Manager.

3- Remember to have a good tuning regarding the stuck threads process detection.
Indeed, if WLS already detects when a thread process  becomes a “Stuck” thread, you have to know that the stuck status doesn’t allow to finish the current work or accept a new one.
So It impacts directly and quickly  the performance of the machine.
4- Don’t forget the basics :  Operating System tuning and/or Network tuning

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Middleware Team
Middleware Team