On 2012-05-22, Alan Corey <alan01...@gmail.com> wrote: > Dump/restore can work remarkably like Symantek/Norton Ghost in this > situation. Get one machine as flawless as possible, then do a dump onto a > spare hard drive. Burn it to a DVD if you like. Then restore onto your > target machines.
I would advise against this method, it can work well enough for an initial rollout but later maintenance is tricky. What works better is to keep track of what *steps* are involved to build a system from scratch and building a process to replicate that. Either hand rolled, or there are many tools (cfengine, puppet, ansible, chef, etc). Whichever way you use, siteXX.tgz, install.site and rc.firsttime (see faq 4) can help you bootstrap things as part of the standard OS install. With a process like this it's easier to keep things under a version control system too which automatically gives you a proper change log which can be hugely useful when troubleshooting.
