Re: Linux hangs...

[Telsa Gwynne <hobbit aloss ukuu org uk>]

On Sat, Apr 07, 2001 at 05:04:55PM -0700 or thereabouts, Andy Tai wrote:
> I am having a similar experience, in GNOME, after
> upgrading to the latest releases (1.4).  If I leave
> the machine running GNOME for a few days (with GNOME
> and X Windows controlling the console, of course), the
> machine tends to lock up and requires rebooting.  My
> kernel is Linux 2.4.2ac20. This does not happen before
> with GNOME 1.2.
> 
> The machine log (/var/log/messages) shows, as the last
> thing before the forced reboot, "kernel: Out of
> Memory: Killing process NNNNN (xteevee)" where NNNNN
> is a process id.  Xteevee is a module of the screen
> saver (Xscreensaver).  I don't know if that has
> anything to do with the lock up.

I have heard of problems involving xscreensaver locking up overnight
in specific circumstances. I believe it's also capable of triggering 
X server bugs.

I would try setting the screensaver to something very plain and seeing
whether it happened there. 

> --- Marius Andreiana <mandreiana rdsnet ro> wrote:
> > > problem. Whenever I leave my terminal for some
> > time
> > > without any activity, and then try to do something
> > > terminal does not response to keyboard and mouse.
> > And
> > console or in gnome ?
> > if gnome
> > 	all gnome (X) is frozen ? try ctrl+alt+backspace,
> > should close X
> > else if console
> > 	try ctrl+q, if you hit ctrl+s before and seemd it's
> > frozen
> > else if kde
> > 	who cares ? :)

I think that last was unwarranted :) 

People often can't tell whether it's the kernel or X (or GNOME, or
a full-screen app within it) which has locked up, especially because
X can get so stuck that the control-alt-backspace (kills X) and 
control-alt-F2 (to get to another console) combinations  just won't 
work. If you have another machine connected to it, then ssh'ing (or 
telnet'ing, whatever) in and looking at the output of 'top' and 
killing the culprit works. If you don't, then yuk. You may well be
stick. 

If you -have- the chance to kill stuff via top, two things to know: 
o try to avoid sending X a kill -9 signal. It doesn't like it at all :)
o kill off some of the clients (the programs running within X) first.
If X suddenly starts running properly then, you have a good idea of
what client (program in X) might be triggering it. 

Telsa