UNIX for Advanced & Expert Users

Im "supporting" at least 2500 HP-UX workstations with CAD-related software with the B.11.11 build. I cant say anymore than that because of my companys sligtly paranoid security policy .
The last few days a new problem has arised from nowhere.

The problem is that users gets logged off when the screenlock should have kicked in (after 10 minutes of idletime). It doesnt happen to all the users and we have the same image installed on all Workstations. No discs are full. And the problem seems to be local since users have the same problem if the try the same computer. If they lend a collegues computer, the problem is gone. I dont think its a RAM issue because there are machines with 2 gig RAM wich runs well while machines with 4 gig RAM can start behaving like this.
The users store and fetch all ther modelfiles from a network of servers. The is no function which unmounts idle networkmounts so that is not the issue either. We are using NIS and Selang for useraccessrights and pushing out UNIX-ids to the local machines.


I havent found any kind of errors in the logs that wasnt there before. Anyone who has experrienced this problem before? Anyone got a clue? Im puzzled...
Im open to any suggestion even the spacy ones. Hey, even stupid suggestions are welcomed!

This is bigger issue than one can think because it takes 15-20 minutes for our users to get all the aplications and networkfunctions up to be able to work on the 3D-models.

Someone pliiiiiiiiiiiiiiiiiz

Hi Laoinjo,

When you say log off, is it a clean logoff or is the X server restarted - i.e. could we be looking at a possible crash?

Does the X log give anything away?

Does it happen *every* time the screensaver is supposed to kick in?

Is there anything else setup to kick in periodically - cron jobs etc. ?

Are all the machines at the same patch level? What patches recently went in?

Cheers,

-c

"When you say log off, is it a clean logoff or is the X server restarted - i.e. could we be looking at a possible crash?"
- They get kicked out to the xserver loginscreeen, so the Xserver is still up and running.

"Does the X log give anything away?"
- Not a thing

Does it happen *every* time the screen saver is supposed to kick in?
- No, maybe 1/10-1/5 of the times. I have just found out though, that the problem seems to occour more frequently the longer the idle times are. So it looks like we can rule out the starting period of the screen lock processes.

"Is there anything else setup to kick in periodically - cron jobs etc. ?"
- No periodical jobs. But the Networkdevices are on 100% alert all the time because of the automated communication with the servers and the Selang and NIS-info that gets pushed out to the workstations.


"Are all the machines at the same patch level? What patches recently went in?"
-Yes. All the machines have identical images installed. We dont patch the machines one by one. We always make clean image copys when we have the need to upgrade. If we do, we always upgrade all the workstation at once. This only happens like once every second year. Sometimes there are application updates though, but none the last 4 months

I will have some time to look in to this further tomorrow. I think I will escalate this problem to the linux- and unix-servergroups at my company to see if there might be some network or NIS- related problem. I will be back with more information in a few days.

The problem is out of my hands now.
The clientgroup wich makes the images is on the problem now.
Everyone are puzzled.
I'll post the solution when they solve it (if they solve it).

Is the TMOUT envionment variable set?

The tmout functions is completely removed from the images we install.

It seems right now that we have a problem with "computer-power-pirates".
Other users who run scripts on other users personal machines to check if that machine is availible for heavy calculations. If it is, they run there heavy algorithms remote in order to speed up their own work.

How this causes the user sitting at the host to be logged out is still a mystery though.
Were beginning to think that the problem is due to that the CAD program (Catia) has only one active lib to control the license and a few other configfiles. The problem is if one user have this CAD-program "running" while idle and another user starts running the same program on the same machine the new users files overwright the first users files and what happens is that the first user get kicked out from the Catia-license server.

But why the user also gets kicked out of the unixsystem is still a mystery.

But atleast were on to it now.