Re: Java application cpu utilization spiking ... taking a long timetocancel.

Hi David,

That's pretty much the path I've been going down ... unfortunately
the diagnostics that the classic JVM provides for garbage collection
are pretty light weight.

There's a tool called iDoctor that can take system snapshots. The problem is that you don't necessarily know *when* to take the snapshots. If you try when there is a memory overrun, it often times out. If you're running WAS, you can use some of the monitoring there. Also, verbose gc can be almost your best friend. BTW, even the asynch collector will go to "stop the world" collection at some point, making it appear that the system is hung.

I'm going to ask the customer if they can increase the amount of
memory allocated to the subsystem the application is running in.

Yes, but that's typically just a short term solution. If you are or effectively are running a heavy batch process, you may have to queue requests or otherwise pause/limit straight ahead operations.

Even if it's an immediate end? IIRC, the customer also
reported that the application took a long time to die if
an ENDJOBABN was used.

Yes. And the IBM people I talked with specifically said not to use ENDJOBABN, that it dould actually make things worse. On occasion, ENDSBS worked, but it really depends on the specific memory situation. One thing you can play with is initial heap psize, which actually works as a gc threshold on the AS.400. HTH,


Joe Sam

Joe Sam Shirah - http://www.conceptgo.com
conceptGO - Consulting/Development/Outsourcing
Java Filter Forum: http://www.ibm.com/developerworks/java/
Just the JDBC FAQs: http://www.jguru.com/faq/JDBC
Going International? http://www.jguru.com/faq/I18N
Que Java400? http://www.jguru.com/faq/Java400

----- Original Message ----- From: "David Gibbs" <david@xxxxxxxxxxxx>
To: "Java Programming on and around the iSeries / AS400" <java400-l@xxxxxxxxxxxx>
Sent: Monday, May 18, 2009 10:46 AM
Subject: Re: Java application cpu utilization spiking ... taking a long timetocancel.

Joe Sam Shirah wrote:

Probably garbage collection,

That's pretty much the path I've been going down ... unfortunately the diagnostics that the classic JVM provides for garbage collection are pretty light weight.

I have asked if they can install the 32bit JVM ... but getting system configuration changes is kind of hard.

I went through a similar issue with WAS over the last few months.
The asynch collector can really run into problems if it can't
complete (usually do to disk swap) before another memory allocation
request comes through. It's very similar to thrashing.

I'm going to ask the customer if they can increase the amount of memory allocated to the subsystem the application is running in.

The ENDJOB doesn't appear to take effect because it wants to clean up
the memory before responding.

Even if it's an immediate end? IIRC, the customer also reported that the application took a long time to die if an ENDJOBABN was used.

david

--
IBM i on Power - For when you can't afford to be out of business
--
This is the Java Programming on and around the iSeries / AS400 (JAVA400-L) mailing list
To post a message email: JAVA400-L@xxxxxxxxxxxx
To subscribe, unsubscribe, or change list options,
visit: http://lists.midrange.com/mailman/listinfo/java400-l
or email: JAVA400-L-request@xxxxxxxxxxxx
Before posting, please take a moment to review the archives
at http://archive.midrange.com/java400-l.