[ILUG] RAID 0 disk failure
Conor Daly
conor.daly at oceanfree.net
Tue Oct 29 09:57:06 GMT 2002
On Tue, Oct 29, 2002 at 09:30:09AM +0000 or so it is rumoured hereabouts,
John Madden thought:
> On (29/10/02 04:26), Paul Jakma didst pronounce:
> > (NB: take a backup of your good disks before doing this)
> >
> Is there a way that I can do this? I can't access any of the disks
> individually from the machine they're in at the moment -- the whole 5 or
> whatever number it is of disks are there as one device and I don't think
> I can access them individually. Would it work if I did one disk at a
> time on another machine? Or, can I actually access each disk separately
> on this machine via different devices?
Individually, you can't see any of the data since it's striped across the
disks. For each disk, do 'dd if=/dev/<disk/device> of=<file>'. If
they're on a hardware RAID controller, you might like to move each disk to
another box and use /dev/hdx
> > let the disk cool for a night, boot up and (imho) /dd/ the raw disk
> > to a file, ie dd if=/dev/rd/c0d0 ... - should read it in one
> > sequential go (rather than seeking back and forth through an fs of
> > indeterminate state of fragmentation), hopefully increasing chance
> > of getting data off it.
> >
> If the /home/ device is /dev/rd/c0d1p1, then the raw disk (which would
> be the full RAID) would be /dev/rd/c0d1 -- am I correct here? Do you
> mean to freeze the disk, slot it back into the array and try and dd the
> whole lot to a file? I've googled (briefly) to see if I can access the
> disks individually, but I can't find anything on the topic.
No, that'll try to read the RAID array and may cause seeks on the bad
disk. Do the bad disk _seperately_ with a beginning to end read.
BTW, when buying the replacement disk, BUY TWO! Put one back in to the
RAID 0 array and rebuild. Then convert the array to RAID 3 by adding the
extra disk as the parity disk (you can do this with an existing RAID 0
set). That way, when this happens again, you'll ba able to rebuild...
> BOFH excuse #1: clock speed
^^^^^^^^^^^
I see you got the important stuff back though!
Conor
--
Conor Daly <conor.daly at oceanfree.net>
Domestic Sysadmin :-)
---------------------
Faenor.cod.ie
9:48am up 2 days, 3:29, 0 users, load average: 0.06, 0.10, 0.04
Hobbiton.cod.ie
10:49am up 1 day, 20:33, 2 users, load average: 0.10, 0.03, 0.01
More information about the ILUG
mailing list