XFS check script on boot?

XFS check script on boot?

[Marco Gaiarin]

I've recently suffered an XFS corruption on a remote server (intel,
debian etch, custom 2.6.X kernel): at some point some process hit an
XFS inconsistency on /var, so /var desappear and suddenly machine
refuse to work.

Being a remote machine with no full-knowledge people there, i've
rebooted it, entered in ssh and stopped all services and tasks,
arriving at the point where i can remount /var readonly.
So i was able to xfs_check the partition (that confirmed me the
corruption), but i was not able to unmount /var, so i was forced to use
'-d' options of xfs_repair. That indeed worked. ;)


This was the first tile i hit a xfs filesystem corruption, so i'm
asking why seems there's no /etc/init.d/checkfs.sh-like script that
check and repair XFS filesystem at boot.

Probably doing fully automatically it is a bit too dangerous, but an
approach like 'normal' fsck, eg if filesystem are too corrupt (it need
'-f' option) ask admin password and force to do it by hand, seems to me
simple and effective.


Someone can explain me? Thanks.

-- 
dott. Marco Gaiarin				    GNUPG Key ID: 240A3D66
  Associazione ``La Nostra Famiglia''                http://www.sv.lnf.it/
  Polo FVG  -  Via della Bontà, 7 - 33078  -  San Vito al Tagliamento (PN)
  marco.gaiarin(at)sv.lnf.it	  tel +39-0434-842711  fax +39-0434-842797

Re: XFS check script on boot?

[Jason White]

On Mon, Mar 17, 2008 at 02:48:47PM +0100, Marco Gaiarin wrote:
 
> This was the first tile i hit a xfs filesystem corruption, so i'm
> asking why seems there's no /etc/init.d/checkfs.sh-like script that
> check and repair XFS filesystem at boot.

Because XFS is a journaling file system which can, and should, recover
automatically from an unclean shutdown such as a system crash or power
failure.

If you get XFS corruption after a reboot then you have a hardware or software
problem that shouldn't be happening and which ought to be investigated and
fixed.

The purpose of a journaling file system is to make time-consuming fsck and
similar checks unnecessary after a reboot.

Re: XFS check script on boot?

[Marco Gaiarin]

Mandi! Jason White
  In chel di` si favelave...

> The purpose of a journaling file system is to make time-consuming fsck and
> similar checks unnecessary after a reboot.

I know exactly that, and indeed, i agree.

I'm not speaking about writing a script that check everytime the
filesystem, but sometime (triggered by /forcefsck?)

The problem arise when you have on a remote site people with low
knowledge, and have to instruct them to run the kernel in
init=/bin/bash mode or in initrd 'emergency' mode, spelling all command
by phone.
[ok, a remote controlled KVM switch is on the buy list ;]

With a such boot script, if a fsck was forced, and a corruption founded,
the system ask for root password and an operator find a shell with /
mounted readonly and all other system unmounted (eg, the 'normal' way
of checking eg, for EXT3; apart that for other filesystem a 'simple'
check are ever tried, and password are asked only for serious corruption
that need 'fsck -f').

A good starting point for trying a recovery.

-- 
dott. Marco Gaiarin				    GNUPG Key ID: 240A3D66
  Associazione ``La Nostra Famiglia''                http://www.sv.lnf.it/
  Polo FVG  -  Via della Bontà, 7 - 33078  -  San Vito al Tagliamento (PN)
  marco.gaiarin(at)sv.lnf.it	  tel +39-0434-842711  fax +39-0434-842797