More at More Ruby on Rails

The Rails Initialization Process

This guide explains the internals of the initialization process in Rails. It is an extremely in-depth guide and recommended for advanced Rails developers.

After reading this guide, you will know:

This guide goes through every method call that is required to boot up the Ruby on Rails stack for a default Rails application, explaining each part in detail along the way. For this guide, we will be focusing on what happens when you execute rails server to boot your app.

Paths in this guide are relative to Rails or a Rails application unless otherwise specified.

If you want to follow along while browsing the Rails source code, we recommend that you use the t key binding to open the file finder inside GitHub and find files quickly.

1 Launch!

Let's start to boot and initialize the app. A Rails application is usually started by running rails console or rails server.

1.1 railties/exe/rails

The rails in the command rails server is a ruby executable in your load path. This executable contains the following lines:

version = ">= 0"
load Gem.bin_path('railties', 'rails', version)

If you try out this command in a Rails console, you would see that this loads railties/exe/rails. A part of the file railties/exe/rails.rb has the following code:

require "rails/cli"

The file railties/lib/rails/cli in turn calls Rails::AppLoader.exec_app.

1.2 railties/lib/rails/app_loader.rb

The primary goal of the function exec_app is to execute your app's bin/rails. If the current directory does not have a bin/rails, it will navigate upwards until it finds a bin/rails executable. Thus one can invoke a rails command from anywhere inside a rails application.

For rails server the equivalent of the following command is executed:

$ exec ruby bin/rails server

1.3 bin/rails

This file is as follows:

#!/usr/bin/env ruby
APP_PATH = File.expand_path('../config/application', __dir__)
require_relative '../config/boot'
require 'rails/commands'

The APP_PATH constant will be used later in rails/commands. The config/boot file referenced here is the config/boot.rb file in our application which is responsible for loading Bundler and setting it up.

1.4 config/boot.rb

config/boot.rb contains:

ENV['BUNDLE_GEMFILE'] ||= File.expand_path('../Gemfile', __dir__)

require 'bundler/setup' # Set up gems listed in the Gemfile.

In a standard Rails application, there's a Gemfile which declares all dependencies of the application. config/boot.rb sets ENV['BUNDLE_GEMFILE'] to the location of this file. If the Gemfile exists, then bundler/setup is required. The require is used by Bundler to configure the load path for your Gemfile's dependencies.

A standard Rails application depends on several gems, specifically:

  • actioncable
  • actionmailer
  • actionpack
  • actionview
  • activejob
  • activemodel
  • activerecord
  • activestorage
  • activesupport
  • arel
  • builder
  • bundler
  • erubi
  • i18n
  • mail
  • mime-types
  • rack
  • rack-test
  • rails
  • railties
  • rake
  • sqlite3
  • thor
  • tzinfo

1.5 rails/commands.rb

Once config/boot.rb has finished, the next file that is required is rails/commands, which helps in expanding aliases. In the current case, the ARGV array simply contains server which will be passed over:

require_relative "command"

aliases = {
  "g"  => "generate",
  "d"  => "destroy",
  "c"  => "console",
  "s"  => "server",
  "db" => "dbconsole",
  "r"  => "runner",
  "t"  => "test"

command = ARGV.shift
command = aliases[command] || command

Rails::Command.invoke command, ARGV

If we had used s rather than server, Rails would have used the aliases defined here to find the matching command.

1.6 rails/command.rb

When one types a Rails command, invoke tries to lookup a command for the given namespace and executes the command if found.

If Rails doesn't recognize the command, it hands the reins over to Rake to run a task of the same name.

As shown, Rails::Command displays the help output automatically if the args are empty.

module Rails::Command
  class << self
    def invoke(namespace, args = [], **config)
      namespace = namespace.to_s
      namespace = "help" if namespace.blank? || HELP_MAPPINGS.include?(namespace)
      namespace = "version" if %w( -v --version ).include? namespace

      if command = find_by_namespace(namespace)
        command.perform(namespace, args, config)
        find_by_namespace("rake").perform(namespace, args, config)

With the server command, Rails will further run the following code:

module Rails
  module Command
    class ServerCommand < Base # :nodoc:
      def perform
        set_application_directory! do |server|
          # Require application after server sets environment to propagate
          # the --environment option.
          require APP_PATH

This file will change into the Rails root directory (a path two directories up from APP_PATH which points at config/application.rb), but only if the file isn't found. This then starts up the Rails::Server class.

1.7 actionpack/lib/action_dispatch.rb

Action Dispatch is the routing component of the Rails framework. It adds functionality like routing, session, and common middlewares.

1.8 rails/commands/server/server_command.rb

The Rails::Server class is defined in this file by inheriting from Rack::Server. When is called, this calls the initialize method in rails/commands/server/server_command.rb:

def initialize(*)

Firstly, super is called which calls the initialize method on Rack::Server.

1.9 Rack: lib/rack/server.rb

Rack::Server is responsible for providing a common server interface for all Rack-based applications, which Rails is now a part of.

The initialize method in Rack::Server simply sets a couple of variables:

def initialize(options = nil)
  @options = options
  @app = options[:app] if options && options[:app]

In this case, options will be nil so nothing happens in this method.

After super has finished in Rack::Server, we jump back to rails/commands/server/server_command.rb. At this point, set_environment is called within the context of the Rails::Server object and this method doesn't appear to do much at first glance:

def set_environment
  ENV["RAILS_ENV"] ||= options[:environment]

In fact, the options method here does quite a lot. This method is defined in Rack::Server like this:

def options
  @options ||= parse_options(ARGV)

Then parse_options is defined like this:

def parse_options(args)
  options = default_options

  # Don't evaluate CGI ISINDEX parameters.
  args.clear if ENV.include?("REQUEST_METHOD")

  options.merge! opt_parser.parse!(args)
  options[:config] = ::File.expand_path(options[:config])
  ENV["RACK_ENV"] = options[:environment]

With the default_options set to this:

def default_options
    Port:               ENV.fetch("PORT", 3000).to_i,
    Host:               ENV.fetch("HOST", "localhost").dup,
    DoNotReverseLookup: true,
    environment:        (ENV["RAILS_ENV"] || ENV["RACK_ENV"] || "development").dup,
    daemonize:          false,
    caching:            nil,
    pid:                Options::DEFAULT_PID_PATH,
    restart_cmd:        restart_command)

There is no REQUEST_METHOD key in ENV so we can skip over that line. The next line merges in the options from opt_parser which is defined plainly in Rack::Server:

def opt_parser

The class is defined in Rack::Server, but is overwritten in Rails::Server to take different arguments. Its parse! method looks like this:

def parse!(args)
  args, options = args.dup, {}

  option_parser(options).parse! args

  options[:log_stdout] = options[:daemonize].blank? && (options[:environment] || Rails.env) == "development"
  options[:server]     = args.shift

This method will set up keys for the options which Rails will then be able to use to determine how its server should run. After initialize has finished, we jump back into the server command where APP_PATH (which was set earlier) is required.

1.10 config/application

When require APP_PATH is executed, config/application.rb is loaded (recall that APP_PATH is defined in bin/rails). This file exists in your application and it's free for you to change based on your needs.

1.11 Rails::Server#start

After config/application is loaded, server.start is called. This method is defined like this:

def start
  trap(:INT) { exit }
  log_to_stdout if options[:log_stdout]


  def print_boot_information
    puts "=> Run `rails server -h` for more startup options"

  def create_tmp_directories
    %w(cache pids sockets).each do |dir_to_make|
      FileUtils.mkdir_p(File.join(Rails.root, 'tmp', dir_to_make))

  def setup_dev_caching
    if options[:environment] == "development"

  def log_to_stdout
    wrapped_app # touch the app so the logger is set up

    console =
    console.formatter = Rails.logger.formatter
    console.level = Rails.logger.level

    unless ActiveSupport::Logger.logger_outputs_to?(Rails.logger, STDOUT)

This is where the first output of the Rails initialization happens. This method creates a trap for INT signals, so if you CTRL-C the server, it will exit the process. As we can see from the code here, it will create the tmp/cache, tmp/pids, and tmp/sockets directories. It then enables caching in development if rails server is called with --dev-caching. Finally, it calls wrapped_app which is responsible for creating the Rack app, before creating and assigning an instance of ActiveSupport::Logger.

The super method will call Rack::Server.start which begins its definition like this:

def start &blk
  if options[:warn]
    $-w = true

  if includes = options[:include]

  if library = options[:require]
    require library

  if options[:debug]
    $DEBUG = true
    require 'pp'
    p options[:server]
    pp wrapped_app
    pp app

  check_pid! if options[:pid]

  # Touch the wrapped app, so that the is loaded before
  # daemonization (i.e. before chdir, etc).

  daemonize_app if options[:daemonize]

  write_pid if options[:pid]

  trap(:INT) do
    if server.respond_to?(:shutdown)
  end wrapped_app, options, &blk

The interesting part for a Rails app is the last line, Here we encounter the wrapped_app method again, which this time we're going to explore more (even though it was executed before, and thus memoized by now).

@wrapped_app ||= build_app app

The app method here is defined like so:

def app
  @app ||= options[:builder] ? build_app_from_string : build_app_and_options_from_config
  def build_app_and_options_from_config
    if !::File.exist? options[:config]
      abort "configuration #{options[:config]} not found"

    app, options = Rack::Builder.parse_file(self.options[:config], opt_parser)
    self.options.merge! options

  def build_app_from_string

The options[:config] value defaults to which contains this:

# This file is used by Rack-based servers to start the application.

require_relative 'config/environment'
run <%= app_const %>

The Rack::Builder.parse_file method here takes the content from this file and parses it using this code:

app = new_from_string cfgfile, config


def self.new_from_string(builder_script, file="(rackup)")
  eval " {\n" + builder_script + "\n}.to_app",

The initialize method of Rack::Builder will take the block here and execute it within an instance of Rack::Builder. This is where the majority of the initialization process of Rails happens. The require line for config/environment.rb in is the first to run:

require_relative 'config/environment'

1.12 config/environment.rb

This file is the common file required by (rails server) and Passenger. This is where these two ways to run the server meet; everything before this point has been Rack and Rails setup.

This file begins with requiring config/application.rb:

require_relative 'application'

1.13 config/application.rb

This file requires config/boot.rb:

require_relative 'boot'

But only if it hasn't been required before, which would be the case in rails server but wouldn't be the case with Passenger.

Then the fun begins!

2 Loading Rails

The next line in config/application.rb is:

require 'rails/all'

2.1 railties/lib/rails/all.rb

This file is responsible for requiring all the individual frameworks of Rails:

require "rails"

).each do |railtie|
    require railtie
  rescue LoadError

This is where all the Rails frameworks are loaded and thus made available to the application. We won't go into detail of what happens inside each of those frameworks, but you're encouraged to try and explore them on your own.

For now, just keep in mind that common functionality like Rails engines, I18n and Rails configuration are all being defined here.

2.2 Back to config/environment.rb

The rest of config/application.rb defines the configuration for the Rails::Application which will be used once the application is fully initialized. When config/application.rb has finished loading Rails and defined the application namespace, we go back to config/environment.rb. Here, the application is initialized with Rails.application.initialize!, which is defined in rails/application.rb.

2.3 railties/lib/rails/application.rb

The initialize! method looks like this:

def initialize!(group=:default) #:nodoc:
  raise "Application has been already initialized." if @initialized
  run_initializers(group, self)
  @initialized = true

As you can see, you can only initialize an app once. The initializers are run through the run_initializers method which is defined in railties/lib/rails/initializable.rb:

def run_initializers(group=:default, *args)
  return if instance_variable_defined?(:@ran)
  initializers.tsort_each do |initializer|*args) if initializer.belongs_to?(group)
  @ran = true

The run_initializers code itself is tricky. What Rails is doing here is traversing all the class ancestors looking for those that respond to an initializers method. It then sorts the ancestors by name, and runs them. For example, the Engine class will make all the engines available by providing an initializers method on them.

The Rails::Application class, as defined in railties/lib/rails/application.rb defines bootstrap, railtie, and finisher initializers. The bootstrap initializers prepare the application (like initializing the logger) while the finisher initializers (like building the middleware stack) are run last. The railtie initializers are the initializers which have been defined on the Rails::Application itself and are run between the bootstrap and finishers.

After this is done we go back to Rack::Server.

2.4 Rack: lib/rack/server.rb

Last time we left when the app method was being defined:

def app
  @app ||= options[:builder] ? build_app_from_string : build_app_and_options_from_config
  def build_app_and_options_from_config
    if !::File.exist? options[:config]
      abort "configuration #{options[:config]} not found"

    app, options = Rack::Builder.parse_file(self.options[:config], opt_parser)
    self.options.merge! options

  def build_app_from_string

At this point app is the Rails app itself (a middleware), and what happens next is Rack will call all the provided middlewares:

def build_app(app)
  middleware[options[:environment]].reverse_each do |middleware|
    middleware = if middleware.respond_to?(:call)
    next unless middleware
    klass = middleware.shift
    app =, *middleware)

Remember, build_app was called (by wrapped_app) in the last line of Server#start. Here's how it looked like when we left: wrapped_app, options, &blk

At this point, the implementation of will depend on the server you're using. For example, if you were using Puma, here's what the run method would look like:

  :Host => '',
  :Port => 8080,
  :Threads => '0:16',
  :Verbose => false

def, options = {})
  options = DEFAULT_OPTIONS.merge(options)

  if options[:Verbose]
    app =, STDOUT)

  if options[:environment]
    ENV['RACK_ENV'] = options[:environment].to_s

  server   =
  min, max = options[:Threads].split(':', 2)

  puts "Puma #{::Puma::Const::PUMA_VERSION} starting..."
  puts "* Min threads: #{min}, max threads: #{max}"
  puts "* Environment: #{ENV['RACK_ENV']}"
  puts "* Listening on tcp://#{options[:Host]}:#{options[:Port]}"

  server.add_tcp_listener options[:Host], options[:Port]
  server.min_threads = min
  server.max_threads = max
  yield server if block_given?

  rescue Interrupt
    puts "* Gracefully stopping, waiting for requests to finish"
    puts "* Goodbye!"


We won't dig into the server configuration itself, but this is the last piece of our journey in the Rails initialization process.

This high level overview will help you understand when your code is executed and how, and overall become a better Rails developer. If you still want to know more, the Rails source code itself is probably the best place to go next.


You're encouraged to help improve the quality of this guide.

Please contribute if you see any typos or factual errors. To get started, you can read our documentation contributions section.

You may also find incomplete content or stuff that is not up to date. Please do add any missing documentation for master. Make sure to check Edge Guides first to verify if the issues are already fixed or not on the master branch. Check the Ruby on Rails Guides Guidelines for style and conventions.

If for whatever reason you spot something to fix but cannot patch it yourself, please open an issue.

And last but not least, any kind of discussion regarding Ruby on Rails documentation is very welcome on the rubyonrails-docs mailing list.