CDE logging users out automatically

[Peter Maddison]

I've got a bizarre problem here which I'm not sure where to look for an answer for.
The X server on one of our users screens keeps randomly restarting. Is there anyway of logging what is causing it? Could it be a hardware fault?
The machine is a B160L, 256Mb memory.

[Leif Halvarsson_2]

Hi
Have you checked the files /$HOME/.dt/errorlog (and errorlog.old), /var/dt/Xerrors, and syslog for suspect error messages.

[Peter Maddison]

I didn't know about /var/dt/Xerrors but as with the other files X just dies.

The syslog.log does contain an error from when the X server has restarted and I presume as the user is logging back into CDE:

Sep 24 10:03:06 lntrw196 /usr/dt/bin/rpc.ttdbserver[18660]: _Tt_session::client_
session_init(): ! Tt_status
Sep 24 10:03:13 lntrw196 last message repeated 40 times

Could CDE have got corupted in some way?

[Alex Glennie]

It doesn't sound like h/w : what O/s ?

I'd be poking around in the files already reffered to AND asking my users what they were doing immediately prior to the hang or what they were running ?

Does X / CDE return after the crash out of interest without user intervention ?

try CDE's diagnostic tool : /usr/contrib/bin/X11/dr_dt ?

sounds a bit like a flakey DNS server or similar could be responsible : try adding local hostname to etc/hosts and edit nsswitch.conf to use files first ? any better ?

[Peter Maddison]

The XServer does return automatically without user intervention. From what the user has said there has not been anything in particular that he has been doing just before it crashes.

I've tried the dt_dr scripts, thanks for that. It returned 5 warnings but I can't imagine any of them particularly apply. I've tried out the points it mentioned anyway.

I think I will just try re-igniting the machine if it happens again since that should be the fastest way to ensure the box has a working copy of CDE installed.

[Alex Glennie]

OK ..... add set -x in .dtprofile and re-check $HOME/.dt/startlog and errorlogs ... out of interest anything in $HOME/errorlog.old?

These files record errors for a CDE users session.

Other thoughts : patch levels note latest CDE patches cure some new tooltlak issues ? are all users affected ? How about root ? what shells are we talking and most importantly do the users have NFS $HOME's ?

[Peter Maddison]

This is the only machine I know to have this problem. We have quite a few users and they all use nfs for there home directories.

The only thing in the errorlog.old is a message saying it has lost connection to the X server and is restarting.

I've put the set -x in so if it happens again it will hopefully give me some more information.

[K.Vijayaragavan.]

Is it a problem being faced by all the users including root of this system? If yes, follow point Nos. 1,2&3. If no, follow the point No. 4

1) Ensure that there is no IP address clashing with the IP address of your workstation. (You may disconnect it from network and see CDE works fine? / simply switch off this workstation and try to ping the IP address of this workstation from some other Node in the same network to trace the node that is clashing.)

2)check the file permission of "/var/spool/sockets" and the files and subfolders uder it. all of them must have permission set with "rwxrwxrwx"

3)If there is any Name service swith problem that may also create such problem.
what happens if you issue the command "nslookup .

4) check the file permission of /etc/hosts.
others (other users must have atleast read permission).
Issue the command "chmod o+x /etc/hosts" to give read permission to others.

[K.Vijayaragavan.]

Sorry! Please substitue the command "chmod o+x /etc/hosts" with "chmod o+r /etc/hosts" in my above posted reply.