Bugs item #588241, was opened at 2002-07-29 13:38
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=376685&aid=588241&group_id=22866
Category: JBossServer
Group: None
>Status: Closed
Resolution: Invalid
Priority: 5
Submitted By: Joe Simone (jsimone)
Assigned to: Nobody/Anonymous (nobody)
Summary: Out-of-memory Exception
Initial Comment:
OS: WinXP Pro
JDK ver: 1.3.1_04-b02
This occurs in Both 3.0.0 final and 3.0.1RC1. Each
iteration creates 10 CMP 2.0 entity beans and returns.
So we get the memory exception after creating about
14,510 entities.
Below is the relevant client code and its result:
int totalPasses = 10000;
for (int pass=1; pass <= totalPasses; pass++) {
// Create a bunch of sample data using the event
manager interface ...
messageBox("CREATE SAMPLE
EVENTS", "Pass "+pass+" of "+totalPasses, " ");
eventIds = eventManager.createSampleEvents(10);
}
[java] ------------------------------------------------------
[java] --- CREATE SAMPLE EVENTS
[java] --- Pass 1451 of 10000
[java] --- Current Time: Sun Jul 28 18:14:18 EDT 2002
[java] ------------------------------------------------------
[java] java.rmi.ServerException: RemoteException
occurred in server thread; nested exception is:
[java] java.rmi.ServerException: null; nested exception
is:
[java] java.lang.OutOfMemoryError
[java] Embedded Exception
[java] null; nested exception is:
[java] java.lang.OutOfMemoryError; nested exception is:
[java] javax.ejb.TransactionRolledbackLocalException:
null; nested exception is:
[java] java.lang.OutOfMemoryError
[java] Embedded Exception
[java] null; nested exception is:
[java] java.lang.OutOfMemoryError
[java] java.rmi.ServerException: null; nested exception
is:
[java] java.lang.OutOfMemoryError
[java] Embedded Exception
[java] null; nested exception is:
[java] java.lang.OutOfMemoryError; nested exception is:
[java] javax.ejb.TransactionRolledbackLocalException:
null; nested exception is:
[java] java.lang.OutOfMemoryError
[java] Embedded Exception
[java] null; nested exception is:
[java] java.lang.OutOfMemoryError
[java] javax.ejb.TransactionRolledbackLocalException:
null; nested exception is:
[java] java.lang.OutOfMemoryError
[java] Embedded Exception
[java] null; nested exception is:
[java] java.lang.OutOfMemoryError
[java] at
sun.rmi.transport.StreamRemoteCall.exceptionReceived
FromServer(StreamRemoteCall.java:240)
[java] at
sun.rmi.transport.StreamRemoteCall.executeCall
(StreamRemoteCall.java:215)
[java] at sun.rmi.server.UnicastRef.invoke
(UnicastRef.java:117)
[java] at
org.jboss.invocation.jrmp.server.JRMPInvoker_Stub.invok
e(Unknown Source)
[java] at
org.jboss.invocation.jrmp.interfaces.JRMPInvokerProxy.i
nvoke(JRMPInvokerProxy.java:128)
[java] at org.jboss.invocation.InvokerInterceptor.invoke
(InvokerInterceptor.java:108)
[java] at org.jboss.proxy.TransactionInterceptor.invoke
(TransactionInterceptor.java:73)
[java] at org.jboss.proxy.SecurityInterceptor.invoke
(SecurityInterceptor.java:76)
[java] at
org.jboss.proxy.ejb.StatelessSessionInterceptor.invoke
(StatelessSessionInterceptor.java:111)
[java] at org.jboss.proxy.ClientContainer.invoke
(ClientContainer.java:76)
[java] at $Proxy1.createSampleEvents(Unknown Source)
[java] at com.highnotes.ebu.client.test.Stress_2.main
(Stress_2.java:91)
[java]
javax.transaction.TransactionRolledbackException: null;
nested exception is:
[java] java.lang.OutOfMemoryError
[java] java.lang.OutOfMemoryError
[java] <<no stack trace available>>
An additional point is that if the entities are immediately
removed after creation, NO memory exception occurs.
In fact, I have had this test case running for almost 36
hours straight with no signs of distress.
Below is the relevant code from the client and its result:
int totalPasses = 10000;
for (int pass=1; pass <= totalPasses; pass++) {
// Create a bunch of sample data using the event
manager interface ...
messageBox("CREATE SAMPLE
EVENTS", "Pass "+pass+" of "+totalPasses, " ");
eventIds = eventManager.createSampleEvents(10);
Iterator iter = eventIds.iterator();
while (iter.hasNext()) {
eventManager.removeEvent((String) iter.next());
}
}
[java] ------------------------------------------------------
[java] --- TEST COMPLETE
[java] --- Pass 10000 of 10000, Total elapsed time: 0
days 0 hrs 26 mins 32 secs
[java] --- Current Time: Sun Jul 28 18:53:08 EDT 2002
[java] ------------------------------------------------------
The session facade has "RequiresNew" as the
transaction attribute and each of the entity bean
methods has "Required".
My expectation is that when the client invokes:
eventIds = eventManager.createSampleEvents(10);
that the createSampleEvents method of the
eventManager Session bean WILL create a new
transaction. Then, that 10 entity beans will be created
under that transaction. Finally, when the
createSampleEvents method returns to the client the
transaction will be complete and a commit will be
performed for the 10 enities created.
I don't see any evidence of this happening.
Shouldn't I be seeing SQL COMMITS in the server.log?
See ref link :
http://www.jboss.org/forums/thread.jsp?
forum=46&thread=18548
----------------------------------------------------------------------
Comment By: Joe Simone (jsimone)
Date: 2002-07-30 10:31
Message:
Logged In: YES
user_id=586242
More data ...
Stopped and started container with max-capacity 100k, reran
test = death
Stopped and started container with max-capacity 50k, reran
test = death at almost the same place
Stopped and started container with max-capacity 20k, reran
test = death at almost the same place
Stopped and started container with max-capacity 10k, reran
test = PASSED after much disk activity as previously noted.
By the way, if I remove the beans after I create a batch of
10, as noted in the initial bug report, then the out-of-memory
condition does not occur. In fact I only found the problem
after taking away the remove in the loop.
Thanks for your help! It is appreciated.
joe
----------------------------------------------------------------------
Comment By: Scott M Stark (starksm)
Date: 2002-07-30 10:29
Message:
Logged In: YES
user_id=175228
So you realize that your test is effectively:
HashMap entityCache = ...
while( iter.hasMore() )
entityCache.put(entity);
and what Bill is telling you is that we will let you create
1,000,000 entities with all their data brought into memory,
so do the math and figure out how much memory you need.
By
dropping the max-capacity you lowering the point at which
the
app server starts paging out the entity cache to disk, just
the same as an OS starts swapping memory to disk when
your
app tries to use too much memory. Read the EJB spec about
the entity life-cycle and caching.
Then go and get the docs subscription to read about how you
can write your own caching policies, optimize loading, etc.
----------------------------------------------------------------------
Comment By: Joe Simone (jsimone)
Date: 2002-07-30 09:45
Message:
Logged In: YES
user_id=586242
Bill,
Thanks for the prompt reply. I lowered the limit from 1 million
to 10 thousand for <max-capacity> in 11 places in the
standardjboss.xml file. I also removed the -Xmx256M from
the run.bat file I had previously put in.
The stress test passed successfull after 20 minutes:
[java] ------------------------------------------------------
[java] --- TEST COMPLETE
[java] --- Pass 10000 of 10000, Total elapsed time: 0
days 0 hrs 20 mins 16 secs
[java] --- Current Time: Tue Jul 30 12:27:22 EDT 2002
[java] ------------------------------------------------------
There was alot of disk activity though, but it did finish. With
the capacity set high I get kinda blip, blip, blip disk activity.
With the capacity set lower I get grind, blip, grind, grind, blip
activity. And it takes much longer, but it passes.
Just to double check I replaced the original
standardjboss.xml that came with 3.0.0 and re-ran the same
exact test after stopping and starting the container fresh and
sure enough Out-of-memory:
[java] ------------------------------------------------------
[java] --- CREATE SAMPLE EVENTS
[java] --- Pass 1438 of 10000
[java] --- Current Time: Tue Jul 30 12:40:29 EDT 2002
[java] ------------------------------------------------------
[java] java.rmi.ServerException: RemoteException
occurred in server thread; nested exception is:
[java] java.rmi.ServerException: null; nested exception
is:
[java] java.lang.OutOfMemoryError
[java] Embedded Exception
[java] null; nested exception is:
[java] java.lang.OutOfMemoryError; nested exception is:
[java] javax.ejb.TransactionRolledbackLocalException:
null; nested exception is:
[java] java.lang.OutOfMemoryError
[java] Embedded Exception
[java] null; nested exception is:
[java] java.lang.OutOfMemoryError
etc. ...
regards,
joe
----------------------------------------------------------------------
Comment By: Bill Burke (patriot1burke)
Date: 2002-07-30 08:24
Message:
Logged In: YES
user_id=176497
The default max cache size for entity beans is 100000, yes, 1
million. Try decreasing it, look in standardjboss.xml. If
decreasing doesn't work, then there is a memory leak either in
JBoss or your application. Let us know what you find.
<cache-policy-conf>
<min-capacity>50</min-capacity>
<max-capacity>1000000</max-capacity>
</cache-policy-conf>
----------------------------------------------------------------------
Comment By: Joe Simone (jsimone)
Date: 2002-07-30 07:36
Message:
Logged In: YES
user_id=586242
By upping the max heap to 256 MB, the problem just
happens later. At pass 5753 (that's 57530 beans created (at
10 per pass)).
I swapped out the DB from DB2 to Hypersonic and the out-of-
memory condition still occurs but sooner for some reason.
Out-of-memory at 1261 passes with HS vs. 1451 passes with
DB2.
joe
----------------------------------------------------------------------
Comment By: Bill Burke (patriot1burke)
Date: 2002-07-29 21:32
Message:
Logged In: YES
user_id=176497
I hope this isn't a braindead answer, but have you tried
increasing the heap size? I think the default entity cache size is
very large and you may be hitting the default heap maz of 64Meg.
-Xmx512M
is the java switch to increase heap.
----------------------------------------------------------------------
Comment By: Joe Simone (jsimone)
Date: 2002-07-29 19:10
Message:
Logged In: YES
user_id=586242
Thanks for responding Luke, I appreciate it.
I am using IBM DB2 Enterprise Edition v. 7.2 with the latest
fixpack 7 applied. I am pretty sure you'll see the same thing
with the freebee personal edition.
Anyway, I'll try and whittle down my application down to just
the pertinent code for the failure and attach here.
I am really hammering the app server hard with long running
tests from multiple clients - or at least trying to - and it hasn't
been holding up for me.
Any help you can render is much appreciated. I am willing to
work with you to make sure this problem is fixed.
regards,
joe
----------------------------------------------------------------------
Comment By: Luke Taylor (luke_t)
Date: 2002-07-29 16:11
Message:
Logged In: YES
user_id=369802
Do you have an example test app that you can post here which
reproduces the problem? The code you have here doesn't show
very much. Also the stack trace from the server log is
generally more useful.
Are you using the internal hypersonic database or an
external DB?
In any case, you won't see any SQL COMMIT commands in the log.
----------------------------------------------------------------------
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=376685&aid=588241&group_id=22866
-------------------------------------------------------
This sf.net email is sponsored by: Dice - The leading online job board
for high-tech professionals. Search and apply for tech jobs today!
http://seeker.dice.com/seeker.epl?rel_code=31
_______________________________________________
Jboss-development mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-development
