[Top] [All Lists]

Re: data recovery after hdd failure

To: thomas <tom@xxxxxxx>, linux-xfs@xxxxxxxxxxx
Subject: Re: data recovery after hdd failure
From: Seth Mos <knuffie@xxxxxxxxx>
Date: Sun, 21 Jul 2002 10:40:38 +0200
In-reply-to: <587043127.20020721003547@xxxxxxx>
Sender: owner-linux-xfs@xxxxxxxxxxx
At 00:35 21-7-2002 +0200, thomas wrote:
so now i got myself a new 80GB and tried to restore at least the XFS
partition (the other being ReiserFS, with less important data).
xfs_repairs reports a bad superblock on hdc2, tries to find an alternate
one and then stops after a while because of an I/O Error. So i tried to

Repairing a disk with IO errors is near impossible.

use dd with conv=noerror, and it seems to work (i stopped after ~500MB).
with xfs_ncheck i get errors and some filelist on that recovered data,

dump the disk with dd to a file on the new disk and repair that instead. This means that although not everything is recoverable you can at least repair what is left of it. Most disks that are failing are able to read sequentially but have big troubling seeking. Older Quantum fireball disks can show you this nicely.

so i guess there is still hope left. i don't know how to use dd properly
though. can i just dd conv=noerror </dev/hdc2 >/dev/hda5 and then
xfs_repair /dev/hda5 or is it recommended to dd to a file?

I recommend doing it to a file but you may have space contraints and the dd file may not be able to fit on there. So dd'ing it straight to another disk might be a option. The geometry will probably have changed which might cause it to do funny things.

I have too much failed IDE disks these days that I only run them in a raid 1 or 5 on important servers. Because IDE disks are rather cheap it's easier to just buy an extra and build me a raid array.

It might just be your lucky day, if you only knew.

<Prev in Thread] Current Thread [Next in Thread>