I've used dd to clone Debian boxes before, but in my case, it was to
identical hardware. You're right in that dd does overwrite the entire
disk, MBR and all, provided that you use correct syntax, of course. :-)|
You have a true cluster-flock of a network. IBM type 2? I haven't
seen that in well over a decade. Nope, there's no way I would even
*think* of trying a multicast over such a network. That thing needs a
total overhaul, new cable plant and all. And unless you're putting
MS-DOG on each of the boxes, the network has nothing to do with that
need for an overhaul.
Yes, your scared-MCSE network admin will most definitely be a problem.
One way to fend off this thorn in your side (and handle your
Linux-leery managerial PHB's) is to point the following out. First
off, to which MS Windows would you convert? You can't get XP anymore,
and there's no way Vista will run on those ancient boxes, so who's
gonna pay for all the new PC's that are "Vista ready"? Second, even if
you *could* get XP, again, who's gonna pay for it?
Here's what I'd do.
1.) Upgrade each of those boxes to *AT LEAST* 512MB DRAM. Sure, you
can run Debian or Ubuntu in 256MB. But for a user experience that's
actually halfway decent, you're not running OpenOffice.org; stick with
KDE 3 and KOffice for optimum use of shared libraries.
2.) Take a look at "preseeding" with Debian/Ubuntu. This is
functionally equivalent to Red Hat's "kickstart" automated-install
system. My understanding is that you cannot use existing partitions,
but since you're cloning disks anyway, that doesn't matter. This way,
you're pretty much guaranteed to get a working GRUB installation.
For EIDE disks (I'm guessing that's what type you have in these boxes),
yep, I always tell the BIOS to use LBA. But if you do that, then make
sure that the box on which you made the original image also had its
BIOS set to use LBA. Things can get a bit dicey at times if you don't
make 'em the same if one's using traditional CHS and the other's using
What is the specific GRUB error that you're seeing? The GRUB
documentation lists 34 possible error codes.
Scott S. wrote:
recent Acer cloning discussion is reminding me of a problem I am having
cloning disks. I am making this a new thread because I do not want to
restore over a network or use compression. This is one of the biggest
and most helpful bunch of Linux geeks there is, thought I would throw
it out here.
Here's the situation. I have a lot of old PCs I want to convert to a
custom Debian/Ubuntu distro. Most of them are using an old Suse
install. The guy who set these up is long gone, and even though these
boxes have worked for years, he otherwise screwed up many things, and
the powers that be are Linux-leery, and want to move to Winderrs.
I do not have Windows machines available for this imaging project, so
that lets out several choices.
The machines are so old, that I think transferring a compressed image
will take more time to clone on uncompressing. The network they are on
is a mish-mash of token ring and ethernet, and it's mostly wired with
old IBM Type 2 wiring. The ethernet ones are using a balun to use the
old wiring. The network is pretty slow, and I don't want to set up
(another) DHCP server for cloning purposes. Many of the machines don't
have name resolution available.
The average hardware is 400-1.700mhz celerons/PIIs typically with 256
meg of ram, all have USB, none have CD drives. They are using a mixture
of network printers, parallel and serial barcode printers.
My custom image (remastersys is great for creating a fresh ISO from an
existing install) is a total of 2.6 gig, one / partition and one swap.
It has the several basic users created, generic printer configurations
set up. It was created on a larger drive with small partitions, then
using dd to only image the first two partitions by block count.
My problem is that most of the machines are too old for either PXE boot
or USB boot. I have experimented with piping dd through netcat
(Knoppix), but it's pretty slow as well. So far what I have been doing
is using Damn Small Linux (it seems to detect the needed hardware
fine). It has a boot floppy that works in the machines well, and then
allows the machine to boot to USB, currently to a USB DVD drive,
hopefully eventually to a syslinux-prepared USB pen drive.
I use http://www.littlesvr.ca/isomaster/ to plop my disk image into the
DSL iso. It lives on a DVD disc. Once I get my floppy-->USB DVD DSL
booted, I just run dd if=/cdrom/masterimage.dd of=/dev/hda bs=64k and
off it goes. Machines modern enough to have USB 2.0 will clone in 20
minutes, older 1.x USB machines take an hour and 20 minutes. I boot
into root, set the default user, hostname and printer, and am done.
There are no recent NV/ATI chipsets to worry about.
While multicast solutions could be much faster on a more modern
network, I am reluctant to try to work with the Windows oriented
network admin to change things. Imaging the machines one at a time is
OK, here's my problem (finally). Sometimes the cloned machine comes up
with a GRUB error in stage 1.5 Sometimes I re-run the imaging process
and it works, sometimes it doesn't. The destination discs are usually
old, but all were in place and working before my imaging process. They
are always larger than my small image. I think my dd method of imaging
should overwrite boot sector and MBR. I haven't yet found anything in
common with the few machines this has failed on.
Is there some quirk about GRUB with drive geometry? It seems like I
read somewhere about large disks in grub should be set in BIOS for
large, or LBA, or something. These are small disks, nothing over 40
K12OSN mailing list
K12OSN redhat com
For more info see <http://www.k12os.org>