ServerLand, centralized server management, enterprise style

[Timo Aaltonen]

On Wed, 14 Dec 2005, Alex Polvi wrote:

> Hello-
>
> After jdub's fantastic email about the new -server stuff, I joined up in
> #ubuntu-server.
>
> Day-to-day I do system administration and love it. What I do not love,
> however, is vendor lock-in. This is one of the reasons we rely on free
> software. *Most* of the place our infrastructure does great with F/OSS.
> The one place it really lacks is with server management -- and I think
> this is a big problem, much bigger then just our organization.
>
> So, I have a solution, and it is here:
>
> https://launchpad.net/distros/ubuntu/+spec/serverland
>
> Anyone else think this is much of a need?  I know I would run it in a
> heatbeat if it exsited in an open form. Do you think others would too?
> What other considerations am I forgetting?

Very much needed.

We have an in-house application for handling the "host detail" part of the 
spec. It is essentially a text file that you edit with $EDITOR
and it is also possible to query stuff like from a database. We have a 
pretty heterogeneous infrastructure here so there are four different 
architectures (alpha, power/ppc, sparc, x86) and few more OSs 
(Tru64, Solaris, AIX, MacOSX, Debian, Ubuntu, RHEL) so it is essential to 
have a single place to hold host (and system) dependant data.

Edited file is committed to a svn repo and the diff mailed to people who 
are interested. Then we have a number of scripts that parse data from this 
db to our NFS-server, installation systems etc. Practically, to have a 
system up-and-running, all you need is to add it to this db and make a 
dns reservation (dns is something that we probably won't integrate to it 
for various reasons, but dhcp for installation system, sure)

I'm not sure if I can publish the scripts for general consumption 
(besides, they contain a lot of specific stuff to our perverse 
infrastructure...), but at least the ideas behind it are free for 
discussion ;)

ps. the backend is meant to be swappable, but the flat text-file approach 
scales well to at least 500 machines, which is approximately the number we 
have atm. (and the file is currently 150kb in size)