RE: Cloning methods
> -----Original Message-----
> From: Eduardo M KALINOWSKI [mailto:eduardo@kalinowski.com.br]
> Sent: Saturday, February 07, 2009 2:06 PM
> Subject: Re: Cloning methods
>
> Nagy Daniel wrote:
> > Hi, again :) :S
> >
> > What's the best method for cloning a partition? [searching for an
> > open-source software alternateive for it :P]
> > I mean cloning like in norton ghost, a program that could "leave"
bad
> > blocks behind, when cloning, and not making a 10 GByte output file
> > [like the partition size is], but just a eg.: a 3 GByte file
[because
> > 3 GByte was used in that partition, all other 7 GByte was free]?
> > [The main goal is to install os on eg.: 15 computers, but only with
> > one physically install in the "reality"]
> >
>
> You may want to take a look at the partimage program.
I will second partimage.
Many years ago I dealt with a cluster that had another OS (I can not
convey my hatred for that system; there simply are not enough words to
do so). As you can guess, we had many problems with it. Anyway, there
was a known issue at the time where a "ghost" process would start on the
nodes (last I checked, still no answer for why; just something in the OS
would suck CPU and memory). The "ghost" process would run the resources
to the ground and blue screen. More often then not this would make the
system blue screen on every startup after and the installation was
unusable. This would mean HOURS of rebuild time as we install the OS,
run updates, restart, install service packs, restart, updates, restarts,
updates, restarts, updates, ect...
It took a lot of work but we eventually setup a partimage server with a
perfect clone of the other OS partition. We then setup the drives on the
nodes with GRUB, the other OS, and a 1GB partition of Linux that would
autorun the partimage script. So whenever we had a node go down, we
reboot the node, select the Linux partition, and then forget about it.
The partimage script would wipe the other OS partition, put a fresh
image on the partition, then reboot into the other OS.
The benefits, besides not wasting a day rebuilding a node, were that we
could update the one image with updates/software and it only took about
1.5 hrs for partimage to run on each of the nodes. This saved us so much
time it isn't even funny. It made dealing with the randomly killed nodes
much less stressful as my day didn't start off with "Oh jeez, I wonder
how many installs will waste away my life today" and replaced it with a
20 minute morning check that consisted of "Blue screen? Reboot. Boot
Linux. Next!"
For this reason, I will always hold a debt of gratitude to partimage. It
is easy to use and the partimage server can be setup to hold a variety
of images and stream the images over the LAN. If you are dealing with a
large amount of systems that you want cloned from a single source, you
should really look into partimage.
Have fun!
~Stack~
PS: No, I don't have access to that cluster anymore. That problem was
not the only one it had. Top 5 maybe, but I don't think it was the
worst. I have many more horror stories on that POS. The first chance we
got we destroyed that waste of time, space, and energy. Before
management could think "well maybe we shouldn't" it was already too late
as we didn't hesitate in wiping it. It was updated with Rocks Cluster
Linux and made useful! :-)
However, if you are in need of help with setting up a similar situation,
let me know. It has been a while, but considering how well I knew the
setup back then, I should be able to recreate it.
Reply to: