🚀 go-pugleaf

Claudio Mettler <claudio.mettler@schweiz.org> wrote:
> Kannst du mit Linux eine fertige win-installation inklusive einem andern quatsch als

I do something else than tar, which works with MS-DOS, OS/2, and the diverse
other MS-* things (inclusive MS-WINDOWS'NT with NTFS).

I simply dump the raw device (/dev/hda, for example) to tape. Restoration
is quite quick. One could also imagine dumping the image to a NFS or
Samba server, or even do a ssh pipe trick.

Something like
   mt -f /dev/st0 setblk 0
   dd if=/dev/hda of=/dev/st0 bs=32768   # backup
   dd if=/dev/st0 of=/dev/hda bs=32768   # restore

On a DDS-3 tape, it takes from 5 to 10 minutes on a 540 MB harddrive.
On a Fast Ethernet, this could well be diminushed to 2 minutes, or
even less with compression.
But, if this is too slow, you can, of course, only backup a partition,
and maybe the first few blocks of the disk so to get the partition table.

Presumably something like:

    dd if=/dev/hda of=/dev/st0 bs=512 count=2
    dd if=/dev/hda2 of=/dev/st0 bs=32768

restored with:
    dd if=/dev/st0 of=/dev/hda bs=512 count=2
    [ maybe here force a reload of /dev/hda's partition table, e.g.
      through fdisk or rmmod/insmod ide ]
    dd if=/dev/st0 of=/dev/hda2 bs=32768   # restore

NB: this one I didn't try.

Now there are drawbacks to the ``raw dump'' method.
   - the new disk (when the old breaks) must have the exact same
     geometry. I had problems with this on older BIOSes when
     changing the motherboard, e.g.
   - if the old disk and the new disks are not good quality IDE
     drives (now, what does that mean a ``quality'' IDE drive), it
     could happen that the bad blocks are not at the same place.
     With SCSI this is not an issue, since the bad blocks are remapped
     internally. I heard that this also happens with ``modern'' IDE
     drives. I never had a problem with this.

I personnally use that for customer simulated operations, in order
to certify hardware product compatibility. I found this to be the
only safe way to ensure I *always* test on the *same exact*
software configuration.

It's so easy to mess OS/2, MS-DOS, MS-* configuration in a way
it's broken or unusable. And there is no easy way to find
differences in configuration.

NB: the university of Geneva has developped an Ethernet card with a PROM
    (probably adapting Etherboot-2.0) which asks, at PC reboot what
    OS you want (Linux, MS-*, etc). When you say so, it *reloads the
    OS* from the server (presumably a raw copy), ensuring a virus-free,
    configuration-hassle free environment on MS-*, and, moreover, the
    ability to easily switch from MS-* to Linux when needed, e.g. by
    courses, or projects.