Zombie Kernel Process on E1 8.11SP1

Hello Gurus, We have a E1 8.11SP1 running for sometime now. Most users are accessing via WebClient while some via FatClient. Weird thing is after some uptime, the Win2003 TaskManager on Ent Server shows a lot of idle JDENET_K, JDENET_N processes consuming Memory space. Have anyone seen something similar before ? Can we set some kernel param so that idle kernel processes are garbage collected ?
Thanks a lot,
Ying-Jie Chen
E1 8.11SP1/Win2003, Tools Rel 8.95I2
Oracle DB 10gR1, Oracle AS 10gR1

I should ask first - do you know they are actually "zombie processes" by viewing the server via SAW or do you just assume they are based on the large number of processes out on the server?

If it is truly zombie processes, ignore this: More than likely you have your "MaxNumberOfProcesses" value set much higher than your "numberOfAutoStartProcesses" value for Type Def6. Also, probably have maxNetProcesses (JDENET) set to a high value. This is going to happen as users connect to the server, over time. I believe they are supposed to spin up kernels as they are needed (for better request load balancing), but in my experience they are started up over time, by request.

You may not need so many, depends on your user load. There are some general rules of thumb, so if you can post your JDE.INI and some information about number of max concurrent users, CNC configuration, etc., that would be helpful.

I don't know of a way to "garbage collect" processes, might want to ask Oracle about that one. In my experience, killing a running process screws things up royally unless they've already been marked as true "zombie" processes (you can check SAW to be certain).

My first and spontaneous though would be to get the font off I2 and move to J1.

Me get burned verdy bad.

Colin

Colin,

Can you elaborate regarding I2? We are seeing some odd things, as well, and are about to boost to J1, but I was curious what specifically you have encountered.

Thanks,
DR

I'm having a problem with P42101 and form W42101. Users will change fields of existing orders and the thin client will just process until a web exception error occurs. Then, the kernel that user is in will just go into a zombie state.

My question is, are your users recieving alot of Web Exception errors? Do you know if there is a certain application that may be hanging? I have a call into Oracle about this. They believe an ESU needs to be applied. I also have several consultants working on the problem as well.

We can now use SAW to identify the zombie kernel process and terminate it manually. The cause most likely is due to that we have quite a lot of Java exceptions from users via WebClient. Many thanks to CMAndersen for elaborating on the JDE kernel processes and the other folks who posted ideas!
Ying-Jie Chen
Systex Corp, Taipei, Taiwan

Did you determine what the problem was with your P42101 application timing out with the web exception error and your kernel zombies? We are having the same problem with P0411. You referenced that you had a case open with Oracle (back in April), but I am not finding any history on the solutions search. Did you have to apply an ESU? Any info you can provide would be much appreciated! Thanks!
E1 8.11 SP1 Tools 8.95_E1