Cognos 8.2 java out of memory error

[LBond]

I built a new server for testing, v8.2, set up the configuration as small configuration, everything is set to default values.  After about 4 people connect, it gets out of memory errors from java.  When I look at java.exe in task manager it just keeps growing (highest I've seen it is 1.5GB).  I tried setting memory from 768 to 1024 in configuration, this just allows cognos to run for a bit longer before getting the out of memory errors.  I reviewed the Java Heap performance proven practice, the Java Garbage Collection tip or technique, and the Cognos 8 Performance Tuning Settings, but I can't see any other changes that I can make to keep this up & running.  I *thought* I had all the setting the same as our prod 8.2 installation, which runs just fine, but apparently not.
Windows 2003 SP2
Dual Core processor
3.75GB RAM

This is the only application running on this server.
Please help.
Lori

[goose]

Hi Lori

Can you post the exact error messages you are getting? Are you running a database instance on this box as well? 4GB may seem like allot of memory but after the OS has used some there may not be enough depending what else is running.

Angus

[ducthcogtechie]

Change the setting to large.
Then manually set the mem value back to 768 instead of 1536.
That's the best it can get (most aggressive garbage collector, with least resource allocation)

If that does not work, you need to take this up with cognos. The jni part running in the java stack
is taking up to much by looping communication or something with your source driver.
Cognos support can help you deciding which issues are known with which driver, and what the next best step is.

[goose]

Oops missed that last bit about cognos being the only app running :D I have only experienced this once before on 8.1, we had to change from Active Directory to LDAP for authentication and that solved the issue.... weird I know.... but that what Cognos support recommended.

Please let us know if you solve this and how...

Thanks

[LBond]

Thanks for the responses.  Yes I have posted this with Cognos support, and I have a test plan to try next week.  For whatever reason, it has settled down & played nice since I posted the issue. 

In my attempts to fix this I did change the setting to Medium (as I had read that Large was unadvisable).  Then I went in and manually changed mem value back to the 768.  So perhaps that is why it has decided to play nice.

Yes, I am using AD for this environment.  We are using LDAP on our old 8.1 install, which is supposed to be upgraded soon (wish me luck).

I will post the resolution if it is something other than the changes to memory stated above.
Lori

[vaisai]

We had the same problem. We used to get the http 500 apache error. Cognos said Tomcat was a third party tool and IIS could be the problem. We freshly installed Cognos on a new server and pointed our Dev box to it. it works ok now. But it still gives the false alarm though the environment is stable. Our version is 8.1.

[eric.yau]

Where did you find this information about "most aggressive garbage collector" when using Large setting?

I've always been trying to figure out the differenece between them.

thanks

Change the setting to large.
Then manually set the mem value back to 768 instead of 1536.
That's the best it can get (most aggressive garbage collector, with least resource allocation)

If that does not work, you need to take this up with cognos. The jni part running in the java stack
is taking up to much by looping communication or something with your source driver.
Cognos support can help you deciding which issues are known with which driver, and what the next best step is.

[ducthcogtechie]

Where did you find this information about "most aggressive garbage collector" when using Large setting?

Internally. I work for cognos.

[LBond]

This server has been running just fine since I changed the configuration to Medium, then manually changed the mem back to 768.  Can't call it a bug, more like a quirk.  Learning more everyday!
Lori