Strange reason for Zombie processes

I just thought I would let everyone in on some strange stuff that we found. We were having some processes go zombie on us on a regular basis. The scheduler kernel, jde_netk , and metadata seemed to be the biggest culprits. We managed to trace it back to loggin off after making a remote connection. If someone logged on to our Enterprise server/database server through remote desktop using the /console mode, or using Dameware, some processes crashed when they logged off. The problem doesn't come up with remote desktop when not using console mode. I didn't dig into this too much after we figured out what was causing it, we just stopped connecting that way. I know that others have recommended using the console mode, so I don't know if it is specific to our setup or what. Our environment has Windows 2003 servers, oracle databsaes, E812, TR 8.96.1.4. I hope this is usefull to someone.

Dave

Hi Dave,

Check this sar 8741284, it sure has the same symptoms as yours but it is reported on tools 8.96.3.2 though. It has a workaround.

Please post your feedback, highly appreciated.

regards.

Kentoy,

Thanks for the information on that SAR. To be honest, I probably won't make that change, I'll just connect through a regular remote desktop session. It doesn't look like it is an issue anymore in 8.97, and we should be going to that soon. I think I'll just wait it out.

Dave

I am having this same problem with 8.97.

This is supposed to be fixed by adding the -Xrs JVM option to the SM configuration.