Re: NetBSD disk got bad sectors. What is the best course of action?

[Stephen Borrill <netbsd%precedence.co.uk@localhost>]

On Wed, 3 Jul 2013, Christos Zoulas wrote:
> In article <20130703131324.GA12493%SDF.ORG@localhost>,
> Mayuresh  <mayuresh%acm.org@localhost> wrote:
> > Seems there is physical damage to the disk.
> >
> > Did fsck -f using another NetBSD system.
> >
> > It shows presence of bad sectors.
> >
> > Backed up the disk.
> >
> > Now shall I try a reinstall or format and reinstall? Are there any other
> > diagnostic tools?
>
> A *long* time ago, when disk got back sectors you could revive them
> for a while by doing a low level format, or by marking specific
> blocks as bad (badsect)... These days disk controllers handle bad
> sectors automatically by reallocating the sectors that got bad from
> a set of reserved sectors. When the disk starts reporting bad
> sectors, this means that the reserved sector list has been exhausted
> and there is no more spares. I would not risk putting more data or
> trying to repair that disk. Unless you are of course either
> emotionally or financially attached to the drive. Depending on if
> it is scsi you can try format and defect management (scsictl), if
> it is ata it probably has SMART and you can check the status (atactl).

I may be wrong, but I think the swap to a reserved sector only happens on 
a write operation. So reads will fail until you attempt to write back at 
which point future reads will succeed. With RAIDframe, the initial read 
failure fails the whole component and no writeback is automatically 
attempted (too bad as it could just copy from the rest of the array).

So if you know failed sector numbers, you could reallocate them by using 
dd if=/dev/zero seek=N count=1 (where N is the sector number) and then 
re-fsck.

If using RAIDframe, usually just the rebuild of the RAID array with -R is 
sufficient.

--
Stephen