sounds like your hard disks are dying (I'm not sure about this, though).
You should really backup your data regularly, anyhow, it is too late
now, but you should do so in future.
If you really need your data, it seems to be a good choice to pay
someone with experience in this area to rescue your data. If you really
want to try this yourself (doing things wrong further worsens the
situation, you have been warned), here is a quick summary of a possible
way of proceeding:
1. Check the self monitoring of the hard disks using 'smartctl -a
/dev/whatever' and if there is anything suspicious, turn the server
off immediately. If your server is already turned off, check the
hard disk's self monitoring after step 5.
2. Buy an external hard drive that is larger than your current ones.
3. Go to http://grml.org/download/ and download the grml rescue/live
4. Burn the image to a CD-ROM or create a bootable USB device
containing the image.
5. Boot your server using this CD-ROM, using 'forensic' as boot
parameter. This boot parameter is important to avoid writing to the
possibly dying harddisks.
6. Try to create a image of the old harddisks on the new harddisk.
http://www.forensicswiki.org/wiki/Ddrescue describes a tool that is
able to do such things (should be installed on the grml CD by default).
If you mix up device numbers, which could be different now, you
could overwrite the data you want to rescue, so be careful.
7. Turn the server off, since you continue rescuing on a different
8. Create an additional copy of the images you created. Everything you
do from now on must happen on the copies to avoid changing the image
9. Now you can do things like running fsck, mounting the image and so
on. To access partitions on the image you can use kpartx (if you
created images of the partitions instead of the hard disk you don't
need kpartx, but creating an image of the hard disks is less error
Since you never did something like this before, you should experiment
with the tools before doing what I described above, or even better: let
someone with experience in this area do it ...
- From: Ploutarchos Michaelides <firstname.lastname@example.org>
- Prev by Date:
- Next by Date:
- Previous by thread:
- Next by thread: