Re: Why doesn't background fsck work ?

> From: Juan Rodriguez Hervella <jrh_at_it.uc3m.es>
> Date: Wed, 18 Jun 2003 17:51:00 +0200
> Sender: owner-freebsd-current_at_freebsd.org
> 
> On Wednesday 18 June 2003 17:39, Bernd Walter wrote:
> > On Wed, Jun 18, 2003 at 04:54:44PM +0200, Juan Rodriguez Hervella wrote:
> > > Hello!:
> > >
> > > I tried to make my Nvidia video card work yesterday, and everytime
> > > I launched the X system my computer hang up. So I had to
> > > make a hard reboot.
> > > (I think I will fix the Xs problem this night at home)
> > >
> > > But I've got another weird problem. :)
> > >
> > > The fsck of my partitions is always made on the foregroud,
> > > although I've heard about something like a delayed/background
> > > file system checker.
> > >
> > > Why is it always made on the foreground ?
> >
> > No softupdates enabled?
> 
> yes, you've got it !
> 
> I haven't got softupdates enabled, but I didn't want to enable it,
> because I've heard that it isn't 100% reliable and I didn't want to lose
> data....

Theer have been no problems with softupdates in regard to data
integrity in either 5.0 or 5.1 release. I do recall a couple of
glitches at various times in current, probably prior to 5.0-Release.

There is an issue of combining softupdate and write cache. In the
event of a power failure, this could lead to a loss of data integrity.
But write-cache is dangerous even without softupdate.

In addition, because of the nature of softupdate, it is possible that
a file is created/updated shortly before a system failure. While data
corruption should not be an issue, the file may simply not be there
after the system re-boots or may be in an earlier (but still
consistent) state.

You should probably turn off write cache on any system that you think
might lose power. (And just how reliable is that UPS? Batteries (DC)
are probably much safer.) But I don't think soft updates are a
significant issue. (Others can feel free to tell me I'm an idiot.)

You should probably read Kirk's Usnix paper on soft updates. It's very
good.
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman_at_es.net			Phone: +1 510 486-8634