Heap Dump ?'s

Heap Dump ?\'s

Hey Everyone...
I'm fairly new to the whole WebSphere and HTML scene but am wondering if there is a way to determine what or who caused heap dumps?

Also, when they are occurring is there any way to stop the session so they don't continue happening?

For instance, I had 5 of them occur this morning and would like to learn who or what caused them so we can talk to the user or person who caused it.
Thanks much!

E811 SP1 Tools Release 8.95N1
WebSphere Version 5 running on AIX Unix platform

Re: Heap Dump ?\'s

I don't know a lot about heapdumps. We're still working through some problems on our end.

I've attached a few documents I had gotten from PeopleSoft support on the subject (yes they are old). They explain more about the possible issue and some ways to troubleshoot it. IBM might be able to help with the troubleshooting. The iConsortium recently learned how to troubleshoot these as well.

Let me know if you have any success.

Regards,
Doug

Re: Heap Dump ?\'s

Diane,

Give me a call. I can help you with this.

Dan

Re: Heap Dump ?\'s

we are experiencing heavy heap dumps at a client as well. We are following Oracle's suggestions on this but still we are not satisfied with the current situation.

Our client works with large grids and oracle tells us not to use the 'go to end' button too much. But if you don't go to the end of the grid, then not all rows are processed? Anyone have a solution for that?

Re: Heap Dump ?\'s

The heap dump issue is pretty straight forward:

1. Get on Tools Release 8.96.1.4 or higher. SAR 8488485 helps clean up objects that have gone out of scope. This is specifically for grid rows which are not freed from memory. Since the memory is not freed you will (no matter what you do) run out of memory.

2. Save 30% of the Heap for large objects using the -Xloratio0.3 (need SDK 1.4.2 --> WAS 6). This in itself is the "SILVER BULLET" that will stop your heap dumps bu you need to be on WAS 6. Check the following URL
http://www-1.ibm.com/support/docview.wss?uid=swg21236509

3. Change the -kCluster size (see Soliution ID 201018061)

4. Change the JDBJ.INI to reduce the amount of caching. See Solution ID 201022163

5. Update the JDK and the Plugin. On WAS the policy is that you can use whatever the latest plugin or JDK is so do it!. Many of your issues will magically disappear.

6. Use the latest version of the proper JDBC driver. This means for a SQL Server database you're using the SQL Server 2005 JDBC Drivers. On an Oracle database download the latest JDBC driver off the web site. For DB2 get the latest fix pack. Finally for AS/400 (iSeries) you MUST use the JDBC driver that comes from the AS/400 and NOT the one from client access. On the iSeries the JDBC drive is update with every CUME pack or Group Database PTF which means if you have an Intel Web server you need to update the driver each time you do a CUME ot Group Database PTF!.

7. Do not use the /3GB switch on Intel Web Servers! JDE only supports W2K3 SP1 and not SP2. The /3GB switch works fine on SP2 but SP1 has too many issues to mention.

Follow all this and no more heap dumps!

Re: Heap Dump ?\'s

Hi Colin

thanks for all the tips! Most of these were never mentioned by Oracle...
All our heapdumps appeared after going to WAS6 and TR8.96.1.4 so those suggestions should help me

Did you receive these from Oracle?