Posted by flatfoot on Wed 8 Mar 2006 at 08:31
These days I have to deploy an environment for a team of developers, some of them prefer stable (Sarge), some can't live without packages from unstable (Sid). I intend to use Sarge (stable) for a server itself, so there's no problem to provide an nfsroot of it. But how is this possible to hit both targets?
Let's pretend that I just manually copy the whole three of an installed unstable and then configure dhcpd/tftpd to point clients to it as nfsroot with an additional selection in pxe boot menu. And then i'll have to spend nights updating it from one of the clients running unstable. Is there any more elegant solution to my problem?
LTSP and other diskless clones are out of business, as client machines are fast enough to do their tasks locally and we don't have a powerful enough server to run multiple instances of Eclipse, OpenOffice and anything else a team could require.
LTSP's design for that task is simply terrible, as it was never meant to directly run local tasks, neither directly access local devices such as cdrom. But we'd love to use locally attached hard disks on a client side for an additional storage and swap.
DRBL doesn't suite much, as it does not support unstable, and Steven Shiau himself has said to me:
It's not easy to do that by the scripts drblsrv and drblpush from DRBL, however, you still can use the diskless package released by Debian to do that.
I've tried diskless twice with no success. Probably because of unstable. Any other thoughts? I'd be glad to see someone come up with a shiny bright idea, as I am simply refuse to give up.
Right now I'm lookin in a Linux VServer or even Xen direction. I know it's simply grotesque to use any of them for such a case, but I see no other options. Does anyone?
This article can be found online at the Debian Administration website at the following bookmarkable URL:
This article is copyright 2006 flatfoot - please ask for permission to republish or translate.