Ruby Mad Science

On a Mac? Here’s how to deploy a test app to Vagrant.

gem install madscience
rvmsudo madscience construct  # install, clone repo, deploy server

This will install all the tools you need locally, clone a repository called “madscience” under the current directory and deploy a test Rails application to that virtual server – by default, on a VM running on your development system.

Want to run your own app instead?

Edit nodes/all_nodes.json.erb in that repo you cloned and it will clone your own real Rails (or Rack) app instead. You’ll need to either use a public URL (like a GitHub HTTP URL) or authorize the SSH key from ~/.deploy_credentials/ to check out your repository.

After that, here’s the same thing to real production, run from inside that “madscience” directory:

vagrant up --provision --provider=digital_ocean

If you edited all_nodes.json.erb to use your app, and ~/.deploy_credentials/digital_ocean.json to use your DO account… Then your app is now running on Digital Ocean. For real.

A scientist with a steaming flask

Want to get started?

Here’s the GitHub repo.

(Worried about the license? It’s MIT.)

Why Not Just Use Chef or Ansible or Whatever For Myself?

Ever set those up? There are a lot of steps and the debugging sucks. This uses Chef under the hood, so you’re getting that, but with the first few days of suck handled already.

Ruby Mad Science uses the right tools - Chef, Capistrano, Vagrant and Librarian as deploy tools, and NGinX, RVM, Unicorn and more to run your app. You could set all that up, or you could let Ruby Mad Science do it. Which one gets you back to coding faster?

It’s a solid stack of tools, and every version is carefully locked down so it keeps working. That takes some doing, even with config management tools like Chef.

It’s secured reasonably. Deploy and provisioning keys are automatically generated and are easy to change. Passwords and SSH keys are stored outside the deploy repo in a credentials directory under your home directory. You could do that, too. But all these things will take time. Want to get back to coding?

Despite all those tools, Ruby Mad Science has a simple top level – a JSON file is used to give the Chef runlist, the set of apps to deploy with Capistrano and (by default) even the third-party cookbooks to set up using the Cheffile. Usually deploy tools are all about repeating yourself. Constantly.

Ruby Mad Science is opinionated but customizable. It’s easy to add third-party cookbooks and Capistrano tasks, and you can diverge entirely from the repo you checked out if you need to – there’s not a huge amount of code and it’s basically all in your deploy repo, in source control.

Out of the box, Ruby Mad Science supports MySQL and Postgres, Redis, MemCache and more if you need them for your Rails app (please don’t use SQLite in production.) Infrastructure like MemCache with a standard Chef cookbook can be added in just a few lines. If you need infrastructure that doesn’t have a standard Chef cookbook, write a cookbook and you can use it from your deploy repo.

By default you can give a convenient list of Rails/Rack apps and static sites in simple JSON format and they get installed and just work nicely. Start easy, configure as you go.

Why Not Heroku?

It’s hard to beat Heroku for getting started quick. However, Heroku is expensive and doesn’t always give you the control you need. Also, you have to rely on them as a third-party service instead of picking your own hosting and choosing how your server is installed. For more details: When Should You Not Use Heroku?

Ruby Mad Science is different. It uses normal open source deploy tools. The Ruby Mad Science source is open (and short!) And it’s really easy to modify your repo to do what you want.

Ruby Mad Science is less of an all-in-one solution and more the jumping-off point you want when you start a new project on a server.

How’s the Support?

This is an open-source project. It’s always nice to get a guarantee of more support than that, when you can.

I sell documentation and support for this code as Rails Deploy In An Hour. It’s the class the code came from originally. But I wanted my customers to be able to use the code for all kinds of projects, and that means open source under a permissive license.

What License?

MIT. That means you can use this for open-source and commercial projects without a problem. I’d appreciate if you contribute fixes back, but you’re not required to. It’s up to you.

The GitHub project has issues, accepts pull requests and so on like you’re used to.

What? Who? Why?

My name is Noah Gibbs. I wrote the book Rebuilding Rails awhile back. When I built Rails Deploy In An Hour I figured, “why not make the code open source?”

It’s easy enough to use, after all. And it’d be a shame to restrict it to customers. Plus, people using tools like Chef, Capistrano and Vagrant can really use a quick start instead of one more book telling them how to use each tool and giving already-out-of-date lists of commands to type and file contents to cut and paste. This is a pretty darn quick start.

Misc Credits

The scientist picture is courtesy of San Jose Library, via Flickr.