BEA 10 Admin server out of memory issue

[shobit shrivastava]

Hello Every one ,

Currently , we have migrated the application from BEA 8 to BEA 10 . However , we are facing an out of memory issue . Admin server is running out of memory after 4-5 days , Though managed server keeps running , eventually application never goes down.

Currently we are using following as memory parameters

WL_JMEM=256
WL_JMEM_MIN=256

WL_ADM_JAVA_XMX=512
WL_ADM_JAVA_MAX_PERM=300

WL_SRV_JAVA_XMX=384
WL_SRV_JAVA_MAX_PERM=192


Can any one throw some light on what is the minimum memory specification required by BEA 10 to support any application. Though I completely agree that the memory specification will differ based on the complexity of application but to support a medium level complex application what should be the memory specification.

We have already tried twice to increase Admin server heap size by changing WL_ADM_JAVA_XMX parameters but sooner or later , Admin server goes down. Any help will be deeply appreciated .

Thanks in advance !

[Deepak Bala]

Can you post the out of memory error that you get ?

Also, how did you confirm that changing the parameters actually changed the heap size ?

[shobit shrivastava]

Thanks Deepak for the reply . your response is the quickest as usual . I would like to thank you for the great work in community

We configured memory parameters in WLserver10.var file by defining the variables

WL_JMEM Default -Xmx: parameter for Admin and Managed servers
WL_JMEM_MIN Default -Xms: parameter for Admin and Managed servers
WL_ADM_JAVA_XMX Specific -Xmx: parameter for Admin server
WL_ADM_JAVA_MAX_PERM -XX:MaxPermSize for the Admin server
WL_SRV_JAVA_XMX Specific -Xmx: parameter for the Managed server(s)
WL_SRV_JAVA_MAX_PERM -XX:MaxPermSize for the Managed server(s)


Now as per the definition , we increased the Prem size not the heap size so far . Let me illustrate entire process so that You can understand the whole work around done by us . Intially , we set WL_ADM_JAVA_MAX_PERM =300 , At that moment Admin server was getting out of memory after 3 days . Later, We increased the Prem size to WL_ADM_JAVA_MAX_PERM =512 , Admin server worked fine for 7 days but again went out of memory. I have no idea how much memory size will be required but definately I believe this out of memory error is not happening due to memory leak in application because the same application runs fine at BEA 8 .Perhaps , We need to identify the right memory requirement to support this application.

Please fine below the error log , though it does not contain any exception trace

####<Nov 6, 2010 3:14:54 PM > <Error> <Kernel> <libank3> <vtvadmin> <
[ACTIVE] ExecuteThread: '70' for queue: 'weblogic.kernel.Default
(self-tuning)'> <<WLS Kernel>> <> <> <1289052894431> <BEA-000802>
<ExecuteRequest failed
java.lang.OutOfMemoryError: Java heap space.
java.lang.OutOfMemoryError: Java heap space
>
####<Nov 6, 2010 3:14:54 PM > <Error> <Kernel> <libank3> <vtvadmin> <
[ACTIVE] ExecuteThread: '71' for queue: 'weblogic.kernel.Default
(self-tuning)'> <<WLS Kernel>> <> <> <1289052894432> <BEA-000802>
<ExecuteRequest failed
java.lang.OutOfMemoryError: Java heap space.
java.lang.OutOfMemoryError: Java heap space
>
####<Nov 6, 2010 3:14:54 PM > <Error> <Kernel> <libank3> <vtvadmin> <
[ACTIVE] ExecuteThread: '69' for queue: 'weblogic.kernel.Default
(self-tuning)'> <<WLS Kernel>> <> <> <1289052894431> <BEA-000802>
<ExecuteRequest failed
java.lang.OutOfMemoryError: Java heap space.
java.lang.OutOfMemoryError: Java heap space
>
####<Nov 6, 2010 3:18:24 PM > <Error> <Kernel> <libank3> <vtvadmin> <
[ACTIVE] ExecuteThread: '64' for queue: 'weblogic.kernel.Default
(self-tuning)'> <<WLS Kernel>> <> <> <1289053104810> <BEA-000802>
<ExecuteRequest failed
java.lang.OutOfMemoryError: Java heap space.
java.lang.OutOfMemoryError: Java heap space
>

Please let me know your opinion.

[Deepak Bala]

java.lang.OutOfMemoryError: Java heap space.

Well the OOM seems to be for heap space, but you are increasing perm gen space. So the variable change should really have no effect. The results are probably coincidental.

Perm gen space is used to load JVM classes. An OOM that occurs out of lack of space in this area will specifically mention it in the error log. For example "outofmemoryerror permgen space".

Try to tweak the Xmx parameter instead. That is the one that controls java heap space. If you run the application through a profiler like visualVm, that will also tell you where the memory leak, if there is one, occurs.

[shobit shrivastava]

Thanks again Deepak .You correctly pointed that there is a glitch in approach .However , I did so because I wanted to avoid a permanent heap memory increase.I was under impression that may be Prem gen increase can solve the problem . Nevertheless ,Now i will tweak the WL_JMEM_MIN =512 parameter (this is Xms parameter for Admin server) and monitor the application. Thank you again for help

[Deepak Bala]

Sure. Hopefully that flag solves your problem

[sudipto shekhar]

Sorry to bump up this post

I would like to add to this i.e. a general recommendation from weblogic, is to set the max perm size to 1/4 of the total heap size!

[Karan Kaw]

Read this URL.
http://www.javacodegeeks.com/2014/09/jvm-permgen-where-art-thou.html

It describes various JVM Memory Options.

[German Gonzalez-Morris]

A good approach would be to dump the Heap when OOM.

Use at least JVM Java 1.5.0_07 and add parameter -XX:+HeapDumpOnOutOfMemoryError to JVM.

The use Eclipse MAT to find out the memory leak.