alt.os.linux - Cloning systems via network

Author

Cloning systems via network

Michael Heiming

2002-08-11, 12:23 pm

Hello,

I'm currently preparing installing of some boxes, same hw and
thinking about setting one up and cloning it trough the network.
Only problem, I need a rescue system with NIC sw/modules to get the
network up, netcat for cloning, shouldn't be very hard, but I need
a tool to setup the hw RAID controller, the module and the /dev
entries, to recognize the created partitions. Currently, I'm
looking in modifying Super-Rescue, to fit my needs.
http://www.kernel.org/pub/dist/superrescue/

Seems a bit oversized for my needs, there's even X on the .iso...
However, it's the first rescue system tested today, which is able
to boot a random box, up to a prompt, out of the box...;-)

A better idea could be, using the RAID controller to duplicate the
disks, they are hot-plugable, and automagically rebuilding the
mirror. I could just exchange one disk after another until they are
all the same, this sounds very easy. Unsure if this will work out,
has anyone done something like this?

If anyone has experience/ideas he would like to share, thx, for
replying.

Michael Heiming
--
Remove the +SIGNS case mail bounces.

Michael Perry

2002-08-11, 5:23 pm

On Sun, 11 Aug 2002 10:41:51 -0700, Michael Heiming wrote:

> Hello,
>
> I'm currently preparing installing of some boxes, same hw and
> thinking about setting one up and cloning it trough the network.
> Only problem, I need a rescue system with NIC sw/modules to get the
> network up, netcat for cloning, shouldn't be very hard, but I need
> a tool to setup the hw RAID controller, the module and the /dev
> entries, to recognize the created partitions. Currently, I'm
> looking in modifying Super-Rescue, to fit my needs.
> http://www.kernel.org/pub/dist/superrescue/
>
> Seems a bit oversized for my needs, there's even X on the .iso...
> However, it's the first rescue system tested today, which is able
> to boot a random box, up to a prompt, out of the box...;-)
>
> A better idea could be, using the RAID controller to duplicate the
> disks, they are hot-plugable, and automagically rebuilding the
> mirror. I could just exchange one disk after another until they are
> all the same, this sounds very easy. Unsure if this will work out,
> has anyone done something like this?
>
> If anyone has experience/ideas he would like to share, thx, for
> replying.
>
> Michael Heiming
> --
> Remove the +SIGNS case mail bounces.

Hi-

There are a few ways that we used to do this at work. One way
is this tool we developed at Linuxcare to do custom linux distros.
I am unsure of its status but what it does is attach to a central
"build" server, allow you to specify the type of system or
partitions you wish to install to, and then it does a modified
rsync type of thing to download the build, install it, and then
run lilo on it when done. The project was hosted on sourceforge
and I believe it was called casper (the friendly "ghost"). It was
written by a few friends of mine that worked with me doing custom
linux distributions for a few different projects.

Its a pretty simple process overall; the way I remember it. You may
have to modify or do a bit of work on the original source code
but it seems to me it was pretty dynamic.

Our goal was to be able to replicate a variety of builds or
deploy them in a data center context where each machine was
exactly the same, had no unique requirements, etc.

BTW, the builds are blasted down very quickly and our experience
with using it was that the resulting file system was laid down
very well for our uses internally as well as some external
clients.

--
Michael Perry
mperry@lnxpowered.org

Michael Heiming

2002-08-11, 6:23 pm

Michael Perry
(<D2E893C7C73E78AE.473D25CA81732BEC.0B2F61BF6F016F66@lp.airnews.net> ):

> I am unsure of its status but what it does is attach to a central
> "build" server, allow you to specify the type of system or
> partitions you wish to install to, and then it does a modified
> rsync type of thing to download the build, install it, and then
> run lilo on it when done. The project was hosted on sourceforge
> and I believe it was called casper (the friendly "ghost"). It was
> written by a few friends of mine that worked with me doing custom
> linux distributions for a few different projects.

I'm running an install-server, but those boxes will have some self
compiled sw and run a different distro. One could generate .rpm and
add them through the install server. However there're only a few
boxes to install, cloning should be much easier/faster, case it's
possible to clone the disks with help of the hw RAID
controller...;-)

Thx for your thoughts

Michael Heiming
--
Remove the +SIGNS case mail bounces.

Michael Perry

2002-08-11, 6:23 pm

On Sun, 11 Aug 2002 16:34:07 -0700, Michael Heiming wrote:

> Michael Perry
> (<D2E893C7C73E78AE.473D25CA81732BEC.0B2F61BF6F016F66@lp.airnews.net> ):
>
>> I am unsure of its status but what it does is attach to a central
>> "build" server, allow you to specify the type of system or
>> partitions you wish to install to, and then it does a modified
>> rsync type of thing to download the build, install it, and then
>> run lilo on it when done. The project was hosted on sourceforge
>> and I believe it was called casper (the friendly "ghost"). It was
>> written by a few friends of mine that worked with me doing custom
>> linux distributions for a few different projects.
>
> I'm running an install-server, but those boxes will have some self
> compiled sw and run a different distro. One could generate .rpm and
> add them through the install server. However there're only a few
> boxes to install, cloning should be much easier/faster, case it's
> possible to clone the disks with help of the hw RAID
> controller...;-)
>
> Thx for your thoughts
>
> Michael Heiming
> --
> Remove the +SIGNS case mail bounces.

Right. We were faced with the same thing. Our installation server
was a debian box but we had to deliver a redhat build over the wire.
What we did was create the file system on a mounted partition that we
wanted to install on the target and then the build software deployed
the "build" that we maintained and not the distribution on the
build-server.

The cloning way works very well too actually. We would often do this
trick where we installed the drive and had a "rescue script" which would
run from a floppy diskette or something and make a mirror image clone or
perhaps a clone from a different build onto the target drive. We never
use hardware RAID for this; but I think your description would work very
well.

Our rescue script basically allowed us to boot from a floppy diskette or a
small linux distribution on a CD (like the lnx-bbc), run a program that we
included on the lnx-bbc and it would clone "a" file system or build onto a
drive that was damaged. When we built this solution, we had delivered a
redhat build which was down around 60mb in size and it was fairly easy to
write a smallish scripted delivery process to deploy/clone a build. I
don't recall the exact method we used for the build but we were able to do
the cloning in just a few steps.

--
Michael Perry
mperry@lnxpowered.org

Sponsored Links