heroku-deploy-rails is a short shell script that I use to deploy my Rails projects to Heroku.

Copy deploy.sh into your Rails project, and commit it. Then rename the git remote for your production app to "production":

$ git remote rename heroku production

The reason you need to do this is that deploy.sh will only make backups of your Postgres database when deploying to a remote called "production". If you'd rather call your remote something else, just edit the script.

You'll also want to enable backups on your app, if you haven't already.

$ heroku addons:add pgbackups:auto-month

If you haven't already added a second heroku app to serve as your staging environment, I'd heartily recommend it (but it is optional as far as deploy.sh goes).

It automates two things for me:

1. It takes a backup of the production database before the deployment.
2. It runs migrations after the code has been pushed.

It also shows you which commits you're about to deploy, and gives you the option to confirm/abort.

All the Heroku deployment tools that I've seen to date are written in Ruby. There's nothing wrong with writing jobs that are best handled by a shell script in Ruby, unless you want to hang around while it starts up and want to trawl through a load of unnecessary boilerplate to discover what's really going on.

Some of them are written in Rake. WTF? That's not what Rake is for; it's for resolving dependencies. It's a build tool. Yes, that's right - the clue is in the name - Rake is like Make, which explains why it doesn't support command line arguments or switches. But whatever. If you like smashing yourself in the face with a blunt instrument, be my guest. ;-)

Shell scripting. You've got to love it. Mr Tomayko knows what he's about, so if you don't believe me, believe him instead.