[mdlug] reiserfsck and badblocks

[Aaron Kulkis]

David Relson wrote:
> G'day,
> 
> My backup hard drive is a USB drive formatted with reiserfs.
> It appears to have a problem or two.  Running reiserfsck v3.6.19
> gives the following error messages:
> 
> --- begin messages ---
> bad_directory_item: block 13501589: The directory item [ ... ] has a
> not properly hashed entry (67)
>  
> bad_leaf: block 13501589, item 8: The corrupted item found (...)
> 
> The problem has occurred looks like a hardware problem. 
> 
> ...
> ...
> 
> bread: Cannot read the block (56093157): (Input/output error).
> 
> --- end messages ---
> 
> I've run "debugreiserfs -B file /dev/sdb1" to get the drive's badblocks
> list. This results in an empty file.
> 
> I've run "reiserfsck --badblocks file /dev/sdb1" where "file" is a
> plain text file containing the block numbers shown above.  Following
> this with "debugreiserfs -B ..." should produce an updated (and
> expanded) badblocks list.  However I get the same empty file.
> 
> Am I doing something wrong???  Any suggestions of other things to try ?
>

Yes (although you didn't know it was doing something wrong):
    you chose reiserfs instead of XFS.

In a relatively few power-off situations, I had several problems
with reiserfs handling the situation poorly.

So, when I installed SuSE 10.1 on my laptop to go to Iraq, I
instead used XFS all of my non-root filesystems (/usr, /home,
/var, /opt, /local, and /tmp).  In Iraq, power-outages were
frequent as generator maintenance takes a couple of hours
and the laptop was usually running in my room, unattended --
on a different generator than the one powering the
communications shop.

I never lost a file in all of those numerous times, plus a
few more since I've been back.

I've had similar experience on my desktop machine: XFS has
not lost a file since I converted over.

XFS is the more robust filesystem.  Use it.