[Gllug] /etc/fstab entries

[Mike Brodbelt]

On Thu, 2006-04-13 at 22:03 +0100, Bruce Richardson wrote:
> On Thu, Apr 13, 2006 at 07:18:48PM +0100, Steve wrote:
> > 
> > I am pondering taking the person to one side and suggesting that my
> > entry is to be preferred.
> > 
> > Am I justified?
> 
> Yes.  The person who set it up may be confused about whether journalled
> filesystems need to be fscked or not.  They do.

Ext3 definitely does - I'm not sure about all of the others. I tend to
use XFS, and the man page for fsck.xfs is quite candid about its
operation:-

    fsck.xfs - do nothing, successfully

For XFS, you check with xfs_check, and repair with xfs_repair, but the
documentation does not suggest any kind of regular checking, with the
xfs_check man page pointing out that it is normally only used if there's
a reason to suspect a problem. So, the XFS guys seem happy to say to
never need to bother with preventative fsck type maintenance, unless
there's something visibly wrong.

My experience with XFS has pretty much borne this out - I've only had
problems on a couple of occasions, and they've tend to be obvious. I've
had the filesystem just vanish, and a log message from the kernel driver
saying something like "I don't like what just happening, so I'm stopping
the filesystem now, go repair it".

JFS and Reiser seem to have a different approach, and their respective
fsck utilities do appear to actually do something, but it's probably
only reasonable to consider the value of fs_passno in the context of the
filesystem on the disk. For XFS at least, having all zeros makes no
difference, and is arguably preferable, as it's not communicating false
information to the casual reader.

Mike

-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug