[NBLUG/talk] hard drive death, filesystem types

[Eric Eisenhart]

On Thu, Dec 16, 2004 at 08:43:17PM -0800, Bob Blick wrote:
> Basically reiserfsck quits when it gets an i/o error, so it won't 
> even test the disk completely, so there's no fixing it either.

That makes life a lot harder.

Easiest, of course, would be to dd the entire partition to another disk and
then check it on that.  This also protects your data if the problem with the
drive is the type that spreads.  (if, say, the problem is a speck of dust
scratching up a platter, reading and writing multiple times will just make
things worse and worse)

However, failing that, I think you can do something like this (probably
while running knoppix or something like it):
badblocks -o hda3.badblocks /dev/hda3
reiserfsck -B hda3.badblocks --fix-fixable --rebuild-tree /dev/hda3

You might need to include "-b 8192" in the badblocks options; but with the
actual block size on the reiserfs filesystem.

Unfortunately, the simple reality is that when blocks on a drive fail, any
data on those blocks is lost.  Whether or not the data on still good blocks
can be recovered usefully is hit-or-miss with any filesystem.
-- 
Eric Eisenhart
NBLUG Co-Founder & Scribe
The North Bay Linux Users Group
http://nblug.org/
eric at nblug.org, IRC: Freiheit at freenode, AIM: falschfreiheit, ICQ: 48217244