I have this exact same problem on the project I'm currently working on, except in our case its a /etc/sudoers file. There are about 160-180 machines that all have pretty much different versions. Since I'm sure it'll turn out to be more than just this one file (actually, its 2-3 already), I ended up writing a service that runs periodically on all these machines that makes secure (https) requests back to a central server (you can use a cacert to verify you are talking to your server) for a list of updated configuration files, then downloads and installs them. The server generates a custom XML file for each node with information about the updates for that particular node, including where to download the file, what permissions/ownership it should have, and what commands to run pre/post install. I have one RPM that I can install on all these machines to keep their config files up to date. There is a good chance this software will be released under some open source liscense, so if your interested, I can keep you up to date. Aaron > I have a similar question I've been meaning to ask. I've > read somewhere that Duke uses yum to completely maintain > machines, which I assume means also maintaining config files. > How is this done, when the config files differ per-machine? > > For one example, I have a large number of machines that have > postfix installed. Each machine has a different main.cf > config file. On my yum repository I can determine what I > want this file to be, but how do I get this into a file on > each client? Should I create a little RPM that depends on > postfix (so it runs after postfix is installed), and have it > run a script that generates main.cf? What do I do if the > config file needs to change? How do I re-run the script in > this RPM? I'm looking for what the best practice here might be. > > Thanks in advance. > > Eric. >
