We do imaging bases and then use scripted installs to “specialize” the machines… although almost all of our setups now aren’t real machines anymore. We are buying uber-beefy opteron setups and running vmware on them so imaging is copying a couple vmdk files and changing the embedded MAC.

We come from a different world though… our machines are all HA frontended and run in pairs at minimum. We hardly ever reinstall a machine… they are almost all upgraded throughout their lifecycle… etc.. so…

Anyways… have fun in the Debian world… just don’t let anyone talk you into Ubuntu for servers yet… that thing is way too green.

The RPM hell argument is about as valid as the Debian is out of date argument these days. I’d even agrue that newer versions of RH based distros have better package management than apt/dpkg.

I can do plenty of hackish cloning. The hardware is pretty much gonna be the same, so I could do images if I wanted to go down that road. I am more a fan of scripted deployment, because it gives much more flexibility than imaging. Be a Debian fan boy all you want, but anaconda is a superior install system over just about anything out there,

I will give you one VERY handy command for people in your situational needs.

mastermachine# dpkg –get-selections | ssh slave1 ‘dpkg –set-selections; apt-get dist-upgrade’

Effectively clones machines (package wise) even if they mismatch architectures and crud like that. Quite handy.