Web Logic Memeory Leaks

Question

I have an application that uses WebLogic Driver for connecting to Database. We use Connection Pooling
We have 11 application threads running and an RMI architecture.
The main server process which delegates the load is given -mx120m
and the other process thread which gets maximum load from the main one is given -mx100m
5 months before the memory was increased in the same way (from -mx72m - -mx100m)
The reason for all this was because the application suddenly hung and when doing a "ps -eaf | grep java" i found that there were around 3 or 4 weblogic threads created and hanging. This never used to happen during the normal cause of running the aplication.
I restart the application & bounce the server..but the application runs for half hour before it hangs again.
Now i increase the memory of the two process as mentioned above it runs perfectly fine. The application is a 12X7 and so it is restarted evey night which should technically clear all memory..
The versions i am using are
Solaris :5.6
Oracle : 8.1.6
Web Logic Driver :3.1
We are trying to get "optmise it" software to check memory leaks.
Any kind of help regarding the cause will be very much appreciated.
Thanks

Answer 1

html tags don't work.
Seems like a lot of mixed up stuff there.<b>It is really getting me confused as there are no
exceptions raised nor where there any indication that
it is a memory leak. All i know is that the
application is hung. It eats up a lot of the Sun
Solaris Memory too...
All the memory comes back when we kill the
application. One more thing is that we have to kill
the processes as it does not respond to nay of the
scripts trying to stop it.</b>By definition if is hung it is generally going to take something extreme to kill it.
>
>>
Typically a memory leak is going to manifest itselfas
a out of memory exception. Unless of course thereis
a bug in the JVM. An you don't mention a out of
memory exception. So either it isn't a memoryproblem
or there is a jvm bug. <b>How do I know whether it is a JVM bug....How can i
find it out</b>Very carefully. Since a lot of people have run the JVM, either it is a known problem, or (more likely) it is in your code (jni.) The way to find it is to reduce the code until the problem goes away. That allows you to identify what process causes it.
For the second you will have
to reduce the problem and then search the bug listand
submit if you don't find it already it. For thefirst
you need to look elsewhere.
the application suddenly hung and when doing a "ps-eaf | grep java" <b>This is done on the command prompt to see if
eveything is running fine.</b>That isn't terribly useful. It tells you that the process is still running. But it doesn't tell you much else, like if it is hung.
>>
You mean like with Runtime.exec()?<b>There i nothing of this sort happening</b>
And you have
threads to handle the output and error streams from
that command? Because if one of those streamsfills
up it will block the thread that spawned it.Threads
solve that. However it doesn't block if the outputis
small enough to fit in the buffer. Because of this
the behavior seems intermittent.<b>Do you mean to ask me how i know that the
application is hung....Well as i told you earlier
there is no indication that abuthing is happning to
the application. It does not take in any transaction.
Does not insert anything into the database.</b>That doesn't mean it is hung. The following loop never will lock up a single threaded app but it is not hung.
while (true) {}
But of course it will mean that not much will get done.
I suspect you are way off course if the only behavior is that is stopped processing and now you are looking for memory leaks.
Have you tried instrumenting your main loops with System.out.println()?
>
We are trying to get "optmise it" software tocheck
memory leaks.
Are you running with an automated profiler, like
OptimizeIt or JProble? If not then how are you
finding memory leaks or performance problems?
<b>I am trying to get the OptimizeIt profiler. but i
am not yet sure what the root cause is</b>
<b><i>There is one more issue of the same kind that
was posted earlier. I am attaching it here and hope
that makes some sense too.
I hava a java application that spawns about 100
threads. Each thread sends a request to WebLogic
server 6.1. We are using
Oracle 8.1.7. They are all
le 8.1.7. They are all running in a Sun box, 2 CPUs
and 6gig memory.
JRE = 1.3.1
SunOS = 5.7
Weblogic = 6.1
Oracle = 8.1.7
(All running on the same box)
When we start the client
en we start the client application, running top shows
that ,
the client JVM process - uses
ent JVM process - uses 40 MB of memory
WebLogic - uses 250 MB of
Logic - uses 250 MB of memory
But the available memory comes
available memory comes down drastically (almost by 350
MB), and on bringing both the client and weblogic
down, we
are not able to get this 350 MB
ble to get this 350 MB back
I would guess that you are misinterpreting something.
For example, I recently had a problem where I thought an application had a memory leak. It turned out that the log file was writing to the same 'mounted' drive that the virtual memory was using. And I was looking at 'free space'. So the free space went down as the log file size went up. It had nothing to do with my application at all.
Running JProbe, doen not show
JProbe, doen not show any memory leak in our
application
we are using the hotspot client
ing the hotspot client version
-Xms = 128m
-Xmx = 128m
</b></i>