On 04.10.00, 13:03:36, Manuel Bouyer wrote:
> On Tue, Oct 03, 2000 at 03:58:51PM -0700, Phil Nelson wrote:
> > 
> > >I've yet to see a
> > >NetBSD system requiring a manual fsck after an unclean shutdown (other than
> > 
> > Well, I have.  I see about 2 a year on my NetBSD/pc532.  It typically
> > happens when I'm doing a build and have a power failure.  It happened
> > a couple of weeks ago on 1.5ALPHA.  It looks like /tmp gets messed up.
> > I can do a "fsck -y /dev/xxx" and have it back.  Most of the power
> > fails come back just fine.  Also, I have not lost anything critical in
> > power fail manual fsck situations.

/tmp shouldn't really hold any data that are meant to survive a panic or
power failure anyway. On most of my systems I have /tmp mounted on an
mfs. (Which is abviously completely lost during reboot / power
failure.)

I sometimes thought that on other systems mounting /tmp asynchronously
might be good idea, but it might prevent an automatic reboot in case
of unexpected inconsistency. Is it advisable or even possible in a
simple manner to have /tmp fsck'd with the -y flag automatically?

> 
> Hum, maybe there's issue with write caching on the disk side ?
> 
> --
> Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
> --
> 

-- 
Bernd Sieker

NetBSD: The Power of Code
		-- Mason Loring Bliss