[svlug] Maintaining config files for various servers

[Mark]

Hi everybody,

This will all be a bit fuzzy, but I don't have a very clear idea of what
exactly I want to do and how. I just thought that a lot of people probably
were in the same situation before:

I am setting up a cluster of servers and want to make this a easily
reproducable as possible. For this reason, I want to keep config files for
all servers in a central place. That way, if I have to re-install a server
for some reason (for example bacause the disk crashed), I can just get the
latest version of all the config files I need from that place.

Some of the files I need are:
iptables, ha (high availablility) config files, ifcfg-eth* files, http
config files, etc.
So they will be spread out over different directories. Is there anything
(preferably an RPM redhat) that does this?

If I had to do it myself, I was thinking about maybe using a directory
structure like:

server1/etc
server1/etc/ha.d
server1/etc/httpd
server1/etc/sysconfig

server2/etc
server2/etc/ha.d
server2/etc/httpd
server2/etc/sysconfig
... 

You get the idea...


Instead of using various subdirectories, I was wondering if it might be
better to only use one per server and naming files accordingly with some
special separator:
etc^ha.d^haresources
etc^sysconfig^iptables

This way I could get all the config files in one scp or http download and
have a script that copies them in place by just replacing "^" with "/". Is
there any problem with using ^ in filenames? I first wanted to use "#", but
that might give me some problems in scripts since it ususally starts a
comment...

Something else I am contemplating at the moment is using yum or up2date to
copy the config files over to the servers, so I only have to change them in
the central repository and can copy them over automatically... Up2date would
be really cool, but it would mean generating RPMs first, which I am still
having some problems with, too....

Also, some of the config files will be identical for all servers, some will
be a little different (just 2 lines out of 100) and others will be
completely different for each machine. How could I avoid too much
redundancy???

It should be some kind of a solution that is easy to maintain, robust and
scalable...

Thanks,

MARK