If you have a single machine that you want to control with Chef (for example, a single home server, or a server that you want to bootstrap to run chef-server for other things), Chef Zero can be a good tool to use. Chef Zero can be used by using the
chef-client -z command. This will treat your local working directory like a Chef server, so you’ll want the following directory structure:
cookbooks/ data_bags/ nodes/ roles/ environments/
Things are optional, but you’ll at least want
Actually, you don’t want cookbooks either.
Well, when you are working with a real chef server, you don’t manually upload cookbooks to it, so why would you do the same with a virtual Chef server? Use Berkshelf like you normally do! Add
cookbooks/ do your .gitignore file, create a separate directory to store your environment cookbook (I use
site-cookbooks/), and make a
Berksfile in the root directory that references it:
source 'https://supermarket.chef.io' cookbook 'server-env', path: './site-cookbooks/server-env'
Now your directory structure will look like this:
site-cookbooks/ server-env/ attributes/ recipes/ .. data_bags/ Berksfile .gitignore
Add data_bags, environments, roles, as necessary. When you are ready, run
berks vendor site-cookbooks and the result will be a
cookbooks/ directory ready for use by chef-zero. Run
chef-client -z and you are ready to converge your server complete with the full power of the chef supermarket in an easy to maintain way. Be sure to check in your
Berksfile.lock so that this is fully reproducible.
The only real downside to this is that while Berksfile sort of works with multiple cookbooks, it can sometimes be awkward to resolve the dependency graph. For example, if you use this as a baseline for multiple servers, and have
server2-env, those cookbooks must have exact version matches on overlapping dependencies from the supermarket. This technique is only suitable for a single environment cookbook.