Re: recovering from a disk failure via raid

["Steven M. Bellovin" <smb%cs.columbia.edu@localhost>]

On Tue, 24 Jun 2008 22:18:37 +0200
Manuel Bouyer <bouyer%antioche.eu.org@localhost> wrote:

> On Tue, Jun 24, 2008 at 11:08:43AM -0400, Steven M. Bellovin wrote:
> > One of two drives I have in a raid1 configuration is showing a read
> > error.  The kernel properly detected this and failed the drive.  The
> > question, of course, is now what?
> > 
> > The first choice, of course, is to install a replacement drive.
> > How do I initiate the recopy?  'raidctl -R'?  (Is it worth just
> > doing that without replacing the drive, on the grounds that
> > rewriting a bad sector will sometimes cause the drive to switch to
> > a spare?)
> 
> It's worth it. I've done it several times with success.
> The read error could also be caused by transient failure ...
> 
I think I'll start that for now, until I actually have the new drives
on hand.  I don't see any downside to trying it, though as someone else
pointed out the drive should now be considered dubious.
> 
> > 
> > But perhaps it doesn't pay to replace the drive -- it's pretty full,
> > and besides, 60GB drives aren't easy to find these days...  If I
> > were to buy two new drives, what is the best way to set up a RAID1
> > on the new pair?  Install one new drive as the sole member of a
> > RAID1 set, install everything on it, then replace the old 60GB
> > drive with the second and initiate reconstruction?
> 
> that would work, but you wouldn't use the extra space on the drive.
> Another way is to setup the second drive as a member of another
> raid-1 in degraded mode, newfs and copy the data the usual way,
> remplace the old drive and initiate a reconstruct. This way you'll
> use the extra space of the new drive.
> 
That was what I meant, actually.


                --Steve Bellovin, http://www.cs.columbia.edu/~smb