You raise some interesting points here and this is a well written post. I am currently building and deploying my first production rails app.

However, something really bothers me about keeping a production config in the repo. Mainly that I don’t believe every developer should have access to production credentials. I plan to implement a method close to the one described in “Agile Web Development with Rails”. I will keep the database.yml out of the repo and put it in a protected directory on my server then use a capistrano hook to copy the database.yml to the prod repo after deployment.

I think following this route will allow me to better secure the production credentials.

Thanks for the suggestion to use raise in lieu of exit. It is valuable to maintain the stack trace. I’ve updated the post (and the environmentalist library at http://github.com/jtrupiano/environmentalist) to reflect this change.

-John

So instead of the unless… exit(1) end sections, use:
raise “Environment directory `#{env_dir}’ does not exist” unless File.directory?(env_dir)
raise “Environment file `#{env_file}’ does not exist” unless File.exists?(env_file)
raise “Database file `#{db_file}’ does not exist” unless File.exists?(db_file)

But who knows, we have maybe 5 developers in my small company, usually all working on different things. I imagine the practices would have to evolve as the size of the team grows.

People raised some interesting points at Bmore on Rails. One I particularly liked was completely leaving SCM off of production servers by doing local checkouts and rsync’ing. Your procedure could fit in here when the local copy is checked out. Another thing to point out is that I’m currently leaving all of my non-production configuration on the production box (shame on me). It’s not simple to restrict a single directory from being checked out (an ignore won’t work, because you want those files in dev). I suppose your process would also aid me here.

I only like my approach slightly better because a) it doesn’t require any changes to the framework or app config files, and b) I’m not checking those files into the same git repo as my application.