How to diagnose spontaneous reboots?

[D. Hugh Redelmeier]

| From: Neil Bird <neil at fnxweb.com>

|   Ordinarily, I'd say that this sort of thing is hardware related, and
| although I will be trying memtest for a while over the weekend, I'm suspicious
| that this has only started happening immediately after I upgraded from Fedora
| 8 to Fedora 10.

Memtest seems like a good idea.

|   I'm running 'nuvexport', which is a script that pulls in MythV recordings
| and spits out normalised video files;  in this case, an XVID AVI using (I
| think) ffmpeg.  Several minutes into the run, the PC just ups and reboots.
| This has now happened twice, and has never before.

Could still be hardware: that might be a heavy CPU load compared to
what you normally do and that might tickle thermal problems.

|   I will also be trying the run from init level 1 just to make sure nothing
| else running might be affecting it.

1?  3 seems less drastic.  Or even just ctrl-alt-f2.

|   But the question is:  if there *is* a bug (kernel?) that's letting a
| user-mode process reboot the machine, how to track it?

If it is a panic, and you are using console mode, you ought to be able
to see what happened.

Alternatively, the traditional technique is to use a serial console.
This has the kernel use the serial port as a console.  For that to
work, you need to have a serial port built into the motherboard (USB
serial port won't do) and you need a "terminal" to attach to the
serial port (these days the terminal is generally just another
computer).

Failing that, you can try kexec/kdump/crash.  In theory this is a great
solution that should work with any machine.  In my experience, hardly
anyone uses this mechanism so it didn't work out of the box in
Fedora 7.  Here are a couple of old bugzilla entries I made:
  https://bugzilla.redhat.com/show_bug.cgi?id=244464
  https://bugzilla.redhat.com/show_bug.cgi?id=244524
See http://fedoraproject.org/wiki/FC6KdumpKexecHowTo
I don't know if things are better now.

Do post your experiences.