Re: which log to diagnose crash

["Richard E. Hawkins Esq." <hawk@eyry.econ.iastate.edu>]

> When X freezes, there's no a priori reason why anything should be logged.
> The rest of the machine should be running normally, and the best course
> of action is to telnet in through the network or a serial line and kill
> the X session (using startx it'll be an xinit process).

This wasn't an X freeze; it was the whole machine. I found it while 
telnetting in; when I got there, X was frozen, and I couldn't get to a 
VC.

> On the other hand, kernel panics are different and may or may not be logged,
> depending on whether it's safe. It will normally tell you if the disks 
> have not been synced (though I've only witnessed that on a VC, not in X),
> in which case you might as well cycle the power.


It didn't sync the disks, so we got to wait while it e2fsck'd a pair of 
4G drives (*sigh*  at least the third disk had been disconnected.  I 
need to poke around and figure out how to get it to do the two scsi 
drives simultaneously . . .).

So how *do* you figure out why a machine crashed?  I've never seen one 
before where I couldn't immediately figure out which hardware did it . 
. .

rick
 
-- 
These opinions will not be those of ISU until it pays my retainer.