Out of memory error. Throughput zero. Server crashes

Question

Hi,
We have this issue about 'java.lang.outofmemory' error, which brings down the
nodes one at a time. The logs don't help at all.
At the time of crash follows:
- Memory shows fine on weblogic console. It cruises flat at approx 160 megs. We
have ms
and mx heap of 512megs.
- jdbc connections on the console show 15 and we have set initial of 15 and max
of 95.
- throughput shows zero on the console
- queue length just spikes and goes like 50+ till the server finally crashes.
- weglogic.log has an entry of jdbc connection leak, which doesn't make sense.
- then a large number of out of memory statements are printed out.
It is our existing application, which worked fine. We put in small additional
features which don't really do any big thing. We have already checked our connection.close()
statements. We are closing all the connections to return them back to the pool.
Our environment is as follows:
OS: Win2k
Weblogic: 6.1 sp4
Clustered with 2 nodes
Any help would be highly appreciated.
Abbas

Answer 1

"siauwea" <[email protected]> wrote in message news:[email protected]...
>
HI,
here i only want to inform that i'm using JRockit java version "1.4.1_03" rather
than jdk.
I wonder is it "-XX:MaxPermSize" cannot be set in JRockit??This option is not available for JRockit - may be it uses different heap structure.
When your server runs out of memory - how much heap is available.
Regards,
Slava Imeshev
"siauwea" <[email protected]> wrote:
Hi,
When i added "-XX:MaxPermSize=128m" in startWeblogic.cmd, why ther server
will
come out "Unknown option: -XX:MaxPermSize=128m"??
Here is what i put in startWeblogic.cmd:
%JAVA_HOME%\bin\java -XX:MaxPermSize=128m %JAVA_VM% %MEM_ARGS% %JAVA_OPTIONS%
-Dweblogic.Name=%SERVER_NAME% -Dweblogic.ProductionModeEnabled=%PRODUCTION_MODE%
-Djava.security.policy="%WL_HOME%\server\lib\weblogic.policy" -DEA_HOME=C:\bea\user_projects\domains\TrekEasy1.0
-DCONF_MODE=SERVER weblogic.Server
Cheers,
Siau Wea
"siauwea" <[email protected]> wrote:
Hi Abbas,
I meet the same problem as you, get the java.lang.OutOfMemory,
you said that u added -XX:MaxPermSize=128m then problem resolved,
but i dont know where should i put this info.
Some say put it in startup parameter, but i dont know where is it.
Can you please tell me clearly which file i should add this line??
Is it in startWeblogic.cmd or other file??
Can you show me all of the setting you did in this jvm option??
Urgent!!
Thanks!!
Siau Wea
"Abbas" <[email protected]> wrote:
Slava,
Thanks for the info. We added -XX:MaxPermSize=128m and it worked. Bingo.
As you said, the parameter MaxPermSize specifies memory for long lived
objects.
Also, what I heard from someone in BEA, that this memory allocationis
something
aside from the heap assigned with -ms and -mx parameters. What I would
really
appreciate is if anyone can talk about any benchmark for this parameter.
The jvm
tuning recommendations for wls 6.1 on BEA's website (http://e-docs.bea.com/wls/docs61/perform/JVMTuning.html#1104200)
doesn't mention this.
Thanks in advance.
Abbas
"Slava Imeshev" <[email protected]> wrote:
Abbas,
Try to add -XMaxPermSize=64M to JVM startup parameters. That will
increase heap size allocaed for long-living objects. If you don't
have
an
application-caused memory leak, this should fix the problem.
Regards,
Slava Imeshev
"Abbas" <[email protected]> wrote in message news:[email protected]...
Hi,
We have this issue about 'java.lang.outofmemory' error, which bringsdown the
nodes one at a time. The logs don't help at all.
At the time of crash follows:
- Memory shows fine on weblogic console. It cruises flat at approx160 megs. We
have ms
and mx heap of 512megs.
- jdbc connections on the console show 15 and we have set initial
of
15 and max
of 95.
- throughput shows zero on the console
- queue length just spikes and goes like 50+ till the server finallycrashes.
- weglogic.log has an entry of jdbc connection leak, which doesn'tmake sense.
- then a large number of out of memory statements are printed out.
It is our existing application, which worked fine. We put in smalladditional
features which don't really do any big thing. We have already checkedour connection.close()
statements. We are closing all the connections to return them backto the pool.
Our environment is as follows:
OS: Win2k
Weblogic: 6.1 sp4
Clustered with 2 nodes
Any help would be highly appreciated.
Abbas