[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Max Mount Count



Ted writes:
> On Tue, Jun 05, 2001 at 02:24:33PM -0400, Robert Dege wrote:
> > /dev/sdb1 has reached maximal mount count, check forced.
> > 
> > Is this normal?  I thought JFS's were supposed to alleviate this check, or
> > is that just for fs corruption?
> 
> The only reason why we do this check at all even for ext2 is paranoia
> about bad IDE cables, cheap PC drives, etc., causing hardware errors,
> and kernel bugs of differing kinds (not necessarily even in filesystem
> code!) that might cause corruption to sneak into a filesystem; it's
> the equivalent of Windows users running Norton Disk Doctor from time
> to time just to make sure nothing bad has slipped in.

FYI for Robert: if you create a new filesystem with mke2fs, it now tells
you what the max_mount_count is, and tells you the filesystem will be
checked every X mounts.  I just changed my e2fsprogs (patch will be sent
to Ted) so that when you run tune2fs to add a journal to the filesystem,
it will also print the message about e2fsck being run periodically and
tune2fs -c/-i to override.

I also changed the e2fsck message so that it tells you how many mounts
triggered the forced fsck.  It _could_ also advertise tune2fs -c/-i so
users would at least know where to look if they want to change this.

> Another trick you can play if you're using LVM and ext3 is to schedule
> a cron job every month or so which takes take a snapshot of the ext3
> filesystem, and then run e2fsck -n on the read-only snapshot in the
> wee hours of the morning.  If the e2fsck turns up any problems, then
> the script can send mail to the administrator informing him that
> there's a problem, and that downtime should be scheduled so that the
> filesystem can be unmounted and fixed.

We need to add the LVM/ext3 integration before we can do this.  In 2.4
the super_operations struct at least has the fields for this, but in
2.2 you would need to make the hooks conditional upon LVM support.  I
will actually have to give this a try before I continue to advocate it
as a solution, although it _does_ sound nice ;-).

Cheers, Andreas
-- 
Andreas Dilger  \ "If a man ate a pound of pasta and a pound of antipasto,
                 \  would they cancel out, leaving him still hungry?"
http://www-mddsp.enel.ucalgary.ca/People/adilger/               -- Dogbert





[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]