Heroku buildpack: Ruby

This is a Heroku buildpack for Ruby, Rack, and Rails apps. It uses Bundler for dependency management.

Usage

Ruby

Example Usage:

$ ls
Gemfile Gemfile.lock

$ heroku create --stack cedar --buildpack https://github.com/heroku/heroku-buildpack-ruby.git

$ git push heroku master
...
-----> Heroku receiving push
-----> Fetching custom buildpack
-----> Ruby app detected
-----> Installing dependencies using Bundler version 1.1.rc
       Running: bundle install --without development:test --path vendor/bundle --deployment
       Fetching gem metadata from http://rubygems.org/..
       Installing rack (1.3.5)
       Using bundler (1.1.rc)
       Your bundle is complete! It was installed into ./vendor/bundle
       Cleaning up the bundler cache.
-----> Discovering process types
       Procfile declares types -> (none)
       Default types for Ruby  -> console, rake

The buildpack will detect your app as Ruby if it has a Gemfile and Gemfile.lock files in the root directory. It will then proceed to run bundle install after setting up the appropriate environment for ruby and Bundler.

Bundler

For non-windows Gemfile.lock files, the --deployment flag will be used. In the case of windows, the Gemfile.lock will be deleted and Bundler will do a full resolve so native gems are handled properly. The vendor/bundle directory is cached between builds to allow for faster bundle install times. bundle clean is used to ensure no stale gems are stored between builds.

Rails 2

Example Usage:

$ ls
app  config  db  doc  Gemfile  Gemfile.lock  lib  log  public  Rakefile  README  script  test  tmp  vendor

$ ls config/environment.rb
config/environment.rb

$ heroku create --stack cedar --buildpack https://github.com/heroku/heroku-buildpack-ruby.git

$ git push heroku master
...
-----> Heroku receiving push
-----> Ruby/Rails app detected
-----> Installing dependencies using Bundler version 1.1.rc
...
-----> Writing config/database.yml to read from DATABASE_URL
-----> Rails plugin injection
       Injecting rails_log_stdout
-----> Discovering process types
       Procfile declares types      -> (none)
       Default types for Ruby/Rails -> console, rake, web, worker

The buildpack will detect your app as a Rails 2 app if it has a environment.rb file in the config directory.

Rails Log STDOUT

A rails_log_stdout is installed by default so Rails' logger will log to STDOUT and picked up by Heroku's logplex.

Auto Injecting Plugins

Any vendored plugin can be stopped from being installed by creating the directory it's installed to in the slug. For instance, to prevent rails_log_stdout plugin from being injected, add vendor/plugins/rails_log_stdout/.gitkeep to your git repo.

Rails 3

Example Usage:

$ ls
app  config  config.ru  db  doc  Gemfile  Gemfile.lock  lib  log  Procfile  public  Rakefile  README  script  tmp  vendor

$ ls config/application.rb
config/application.rb

$ heroku create --stack cedar --buildpack https://github.com/heroku/heroku-buildpack-ruby.git

$ git push heroku master
-----> Heroku receiving push
-----> Ruby/Rails app detected
-----> Installing dependencies using Bundler version 1.1.rc
       Running: bundle install --without development:test --path vendor/bundle --deployment
       ...
-----> Writing config/database.yml to read from DATABASE_URL
-----> Preparing app for Rails asset pipeline
       Running: rake assets:precompile
-----> Rails plugin injection
       Injecting rails_log_stdout
       Injecting rails3_serve_static_assets
-----> Discovering process types
       Procfile declares types      -> web
       Default types for Ruby/Rails -> console, rake, worker

The buildpack will detect your apps as a Rails 3 app if it has an application.rb file in the config directory.

Use em-postgresql-adapter on Heroku

This buildpack is designed to detect and write a database.yml file for em-postgresql-adapter, so that you can apply the "Async-Rails" technique described by Ilya Grigorik at https://github.com/igrigorik/async-rails. Here are steps to do this:

(1) Add these gems to your Gemfile:

gem 'rack-fiber_pool', :require => 'rack/fiber_pool'
gem 'eventmachine', '>= 1.0.0'
gem 'em-postgresql-adapter', :git => 'git://github.com/leftbee/em-postgresql-adapter.git'
gem 'em-synchrony', :git => 'git://github.com/igrigorik/em-synchrony.git',
                    :require => ['em-synchrony',
                                 'em-synchrony/activerecord']
gem 'em-http-request', :require => 'em-http'

(2) Then modify the config.ru file:

use Rack::FiberPool, :size => 100

You can change the pool size.

(3) You can also apply some more advanced configuration for the DB connection. From terminal, login to your Heroku account, the add these env variables:

heroku config:add DATABASE_POOL=20 --app <your-app-name>
heroku config:add DATABASE_CONNECTIONS=5 --app <your-app-name>

Be careful when set these variables because the Heroku Postgres Dev limits only 20 connections pool. See https://devcenter.heroku.com/articles/heroku-postgres-starter-tier#limits.

(4) Use the custom Heroku buildpack:

heroku config:add BUILDPACK_URL=https://github.com/kidlab/heroku-buildpack-ruby --app <your-app-name>

See more at: https://devcenter.heroku.com/articles/buildpacks#using-a-custom-buildpack

You can see another fork of Async-Rails for PostgreSQL at https://github.com/jackquack/async-rails32_postgres

Assets

To enable static assets being served on the dyno, rails3_serve_static_assets is installed by default. If the execjs gem is detected then node.js will be vendored. The assets:precompile rake task will get run if no public/manifest.yml is detected. See this article on how rails 3.1 works on cedar.

Hacking

To use this buildpack, fork it on Github. Push up changes to your fork, then create a test app with --buildpack <your-github-url> and push to it.

To change the vendored binaries for Bundler, Node.js, and rails plugins, use the rake tasks provided by the Rakefile. You'll need an S3-enabled AWS account and a bucket to store your binaries in as well as the vulcan gem to build the binaries on heroku.

For example, you can change the vendored version of Bundler to 1.1.rc.

First you'll need to build a Heroku-compatible version of Node.js:

$ export AWS_ID=xxx AWS_SECRET=yyy S3_BUCKET=zzz
$ s3 create $S3_BUCKET
$ rake gem:install[bundler,1.1.rc]

Open lib/language_pack/ruby.rb in your editor, and change the following line:

BUNDLER_VERSION = "1.1.rc"

Open lib/language_pack/base.rb in your editor, and change the following line:

VENDOR_URL = "https://s3.amazonaws.com/zzz"

Commit and push the changes to your buildpack to your Github fork, then push your sample app to Heroku to test. You should see:

-----> Installing dependencies using Bundler version 1.1.rc

NOTE: You'll need to vendor the plugins, node, Bundler, and libyaml by running the rake tasks for the buildpack to work properly.

Flow

Here's the basic flow of how the buildpack works:

Ruby (Gemfile and Gemfile.lock is detected)

• runs Bundler
• installs binaries
  • installs node if the gem execjs is detected
• runs rake assets:precompile if the rake task is detected

Rack (config.ru is detected)

• everything from Ruby
• sets RACK_ENV=production

Rails 2 (config/environment.rb is detected)

• everything from Rack
• sets RAILS_ENV=production
• install rails 2 plugins
  • rails_log_stdout

Rails 3 (config/application.rb is detected)

• everything from Rails 2
• install rails 3 plugins
  • rails3_server_static_assets