Need help with dying drive/restoring data

[Peter Risdon]

Jonathon McKitrick wrote:
> Hi all,
> 
> Help!  My laptop drive seems to be dying, and while I did keep backups, the
> last one was a bit old.
> 
> When I boot up, the drive makes clanking sounds I've never heard before, and
> never finishes the load.  I'm going to make a rescue disk, but does anyone
> have a strategy for how I could handle the delicate job of getting my
> updated data off the drive without making matters worse?  So far, I figure I
> will boot the rescue disk and try to mount the filesystems.

This is based on experience rather than the extremely detailed knowledge 
of some other posters - like most people, I've had to get data off 
failing disks from time to time and so some strategies have emerged.

One moderately obvious thing - if you boot from a rescue disk, mount the 
damaged drive read only.

The filesystem will probably be marked unclean, and, unfortunately, 
running fsck can make the drive fail again before you have a chance to 
get any data off it. In an extreme case, dd might be your only option.

I've found that attempts to copy/tar/dump the whole filesystem in these 
cases often fail. It's certainly worth trying once, but if it fails you 
can copy/tar/dump parts of the disk individually, starting with the most 
important areas, and you've a fairly good chance of at least partial 
success.

I have a feeling that this helps because it avoids too high a level of 
continuous disk activity. If that's the case, I've started wondering 
whether using rsync with the --bwlimit argument is worth investigating 
as a method of limiting throughput. I haven't tried this, though.

You might find that there are areas of specific damage that have to be 
worked around and these can be identified by a process of elimination.

I also try not to let the drive go *cold* once these problems have 
started developing. A dying disk seems to be more likely to fail 
completely on power up than at any other time. This isn't meant to 
contradict the freezer idea suggested by another poster, which is widely 
recommended and definately worth trying. It means try not to keep 
rebooting once you start to recover the data, if you can manage it.

Peter.