Hi Raja,

Thanks very much for your detailed response.  I will keep your
suggestions in mind.

On Tue, Oct 19, 2010 at 1:42 PM, Raja Subramanian
<[email protected]> wrote:
> On Tue, Oct 19, 2010 at 1:02 PM, Arun Khan <[email protected]> wrote:
>> I have an external WD 1TB hard disk with a XFS file system.  It used
>> to be mounted to my desktop via an e-sata port on the mobo.
> ...
>> Anyway,  I have hooked up the 1TB device to my temporary desktop via
>> USB and the kernel is not reporting any error.
>>
>> However, there are file system errors and it refuses to mount.
>>
>> Using SystemRescueCD I have tried to repair the FS with the following:
>>
>>                            xfs_repair -v -L /dev/sdf1
>
> XFS fsck requires a LOT of RAM to check large file systems.  Please
> ensure you have at least 2GB of RAM on your system and are running
> a 64bit kernel and userland.  It's also good to have adequate swap
> mounted -- you don't want xfs_check to run out of RAM midway.

Yes, I am using 64bit kernel with 8GB RAM.  I am using SystemRescueCD.
  Memory was not the issue; top showed only 670MB in use.

> I always "xfs_check" first to identify errors.  Then try a dry run with
> "xfs_repair -n" before turning xfs_repair loose.
>
>
>> I see no output on the console (even after letting it run for a couple
>> of hours) and I am unable to break out of the session (Cntrl-C).
>

It turns out I was overly optimistic about the condition of the hard
disk.  Although, it showed no disk read errors last night, they showed
up today when I was attempting yet another "xfs_repair" with the -P
option.  This time around I did get some output on the console with it
aborting on disk I/O error.  Time to RMA the disk to WDC.  In this
case,  the files were mostly ISO of various distros so not much loss.


> USB is very slow, so a lot of patience is required :-)  Or hook up the
> HDD to your internal SATA controller and rerun fsck.

I do not have any spare sata ports on the server and esata port either.

> xfs_repair is probably stuck on kernel IO so Control-C is not stopping
> it.  If you don't see any disk activity, it should be safe to reboot the box
> and try the above.
>
>
>> I have also done ddrescue to extract about 150GB of the FS and try to
>> repair the extracted file - no success.
>
> Do you have a full image of the XFS partition?  What is the size of your
> XFS partition.

The partition size is the full disk - around 976GB.  I do not have an
equivalent spare so I tried about 150GB and tried mounting the image
file - did not work.  Clutching at straws before letting it go :(

> If you have LVM setup, suggest that you create a snapshot before running
> xfs_* so you can revert in case things do not go as planned.
>
>
> Not to discourage you but my experience in recovering XFS has mostly met
> with limited success.  xfs_repair renames the files with their inode numbers
> and generally creates a mess.  I've often had to reinstall IRIX after
> a file system
> corruption.

With this experience, I agree with your suggestion re. file system
choice.   I too have had very limited success with XFS repair on
another occasion and ended up reformatting the volume.

-- Arun Khan
_______________________________________________
ILUGC Mailing List:
http://www.ae.iitm.ac.in/mailman/listinfo/ilugc

Reply via email to