[plug] kernel panic tip

[Bernard Blackham]

On Fri, Jan 02, 2004 at 11:32:49AM +0800, Chris Caston wrote:
> Any idea if this brings or can be setup to bring a graceful shutdown?

After a panic, there's no such thing as graceful shutdown. Or any
sort of shutdown other than the power/reset switch.

> I've had a server kernel panic and found the only thing I could do was
> to reset manually afterwhich it came back up and required a lot of
> fscking to clean up all the messed up inodes.

Were you using a journalling fs like XFS, JFS, ext3 or resier? If
you can capture the panic (serial consoles are good if you know in
advance it's going to panic, otherwise digicams suffice) it'll tell
you what process triggered it. If you can get it into a file
somehow and run ksymoops over it, you can find out exactly why it
oopsed, but this is only if you're interested in pouring over kernel
source to tinker with it.

> After a few iterations of this it stop booted and had to be cleanly
> reinstalled.

What filesystem were you using? Some people will claim certain
filesystems are prone to failure if the slightest thing goes wrong,
from bad past experiences.

<tale>
ext3 has never failed me. Even after I mkxfs'd over the top
of an ext3 partition, I could e2fsck and get all crucial files back
(mkxfs on an ext3 partition you wanted is a biiig mistake). Funny
thing was that after that incident, the partition then mounted as
both ext3 *and* XFS (not simultaneously). Playing around later, I
could get about 40MB of data on both an XFS and ext3 filesystem on
the same 500MB partition until the filesystems started getting in
the way of each other. Fun fun :)
</tale>

Bernard.

-- 
 Bernard Blackham <bernard at blackham dot com dot au>