Question about forcing fsck at boottime

[Erik Trulsson]

On Tue, Mar 31, 2009 at 04:04:53PM +0100, Chris Rees wrote:
> On Tue, 31 Mar 2009 16:00:18 +0530
> manish jain <invalid.pointer at gmail.com> wrote:
> > Having bgfsck enabled is like
> > inviting a dragon to dinner when this happens.
> 
> 2009/3/31 RW <rwmaillists at googlemail.com>:
> > If you've done a normal install, soft-updates aren't enabled on /,
> > so it will get foreground checked by default.
> >
> > If I were you I'd reboot into single user mode and do a full fsck on it.
> 
> 
> Seriously, why is everyone against background fsck? Can anyone give a
> good reason? Please?

For background fsck to work as it is supposed to, it is necessary that
only certain errors can occur on the filesystem.  Other types of errors
cannot be corrected by a background fsck.

To make sure that only the allowable errors can occur it is necessary
for soft updates to be used and working as it is supposed to.

For soft updates to work as it is supposed to the disk subsystem must
provide certain guarantees on when and in which order blocks are written.

Normal PATA/SATA disks with write caching enabled (which is the default) do
not provide these guarantees.  Disabling write caching on will make them
adhere to the assumptions that soft updates make, but at the cost of a
severe performance penalty when writing to the disks.


In short therefore on a 'typical' PC you can fairly easily get errors on a
filesystem which background fsck cannot handle.



It is also the case that background fsck relies on snapshots to work,
At least in the past snapshots had stability problems.  Things are supposed
to be better these days, but many people have long memories for these kind
of problems.






-- 
<Insert your favourite quote here.>
Erik Trulsson
ertr1013 at student.uu.se