Skip to content
This repository

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

ActiveRecord connection proxy for master/slave connections

branch: master

Fetching latest commit…

Octocat-spinner-32-eaf2f5

Cannot retrieve the latest commit at this time

Octocat-spinner-32 lib
Octocat-spinner-32 test
Octocat-spinner-32 LICENSE
Octocat-spinner-32 README.markdown
Octocat-spinner-32 init.rb
README.markdown

masochism

Scream

photo by ArtWerk

The masochism plugin provides an easy solution for Ruby on Rails applications to work in a replicated database environment. It works by replacing the connection object accessed by ActiveRecord models by ConnectionProxy that chooses between master and slave when executing queries. Generally all writes go to master.

Quick setup

First, setup your database.yml:

# default configuration (slave)
production: &defaults
  adapter: mysql
  database: app_production
  username: webapp
  password: ********
  host: localhost

# setup for masochism (master)
master_database:
  <<: *defaults
  host: master.example.com

To enable masochism, this call is required:

# enable masochism
ActiveReload::ConnectionProxy.setup!

Example usage for a Rails application:

# in environment.rb
config.after_initialize do
  if Rails.env.production?
    ActiveReload::ConnectionProxy::setup!
  end
end

Considerations

Thinking Sphinx

Thinking Sphinx inspects the connection object to determine the database adapter. Because masochism works by putting the connection proxy in its place, TS will get confused about ActiveReload::ConnectionProxy and abort. A possible workaround is to monkeypatch TS right to hardcode our adapter after masochism has been enabled:

# ConnectionProxy from masochism confuses TS
ThinkingSphinx::Index.class_eval do
  def adapter() :mysql end
end

ThinkingSphinx::AbstractAdapter.class_eval do
  def self.detect(model)
    ThinkingSphinx::MysqlAdapter
  end
end

In newer versions of Thinking Sphinx (observed in 1.1.3), different parts have to be patched:

# for ThinkingSphinx v1.1.3
ThinkingSphinx::AbstractAdapter.class_eval do
  def self.detect(model)
    adapter = model.connection.adapter_name
    case adapter
    when 'MySQL'
      ThinkingSphinx::MysqlAdapter.new(model)
    when 'PostgreSQL'
      ThinkingSphinx::PostgreSQLAdapter.new(model)
    else
      raise "Invalid database adapter #{adapter.inspect}: Sphinx only supports MySQL and PostgreSQL"
    end
  end
end

ThinkingSphinx::Index.class_eval do
  # we have to overwrite this method because it tries to access
  # @model.connection.instance_variable_get(:@config)
  def set_source_database_settings(source)
    config = @model.configurations[Rails.env].symbolize_keys

    source.sql_host = config[:host]           || "localhost"
    source.sql_user = config[:username]       || config[:user] || ""
    source.sql_pass = (config[:password].to_s || "").gsub('#', '\#')
    source.sql_db   = config[:database]
    source.sql_port = config[:port]
    source.sql_sock = config[:socket]
  end
end

Litespeed web server or Phusion Passenger (mod_rails)

If you are using the Litespeed web server or Passenger (mod_rails), child processes are initialized on creation, which means any setup done in an environment file will be effectively ignored. A brief discussion of the problem is posted here.

One solution for Litespeed/Passenger users is to check the connection at your first request and do the setup! call if your connection hasn't been initialized, like:

# in ApplicationController
prepend_before_filter do |controller|
  unless ActiveRecord::Base.connection.is_a? ActiveReload::ConnectionProxy
    ActiveReload::ConnectionProxy.setup!
  end
end

Advanced

The ActiveReload::MasterDatabase and ActiveReload::SlaveDatabase abstract models use the 'master_database' and 'slave_database' settings, respectively.

In 'database.yml' configuration file, both settings can be specified either per-environment or global (like in the first example of this document).

Examples:

login: &login
  adapter: postgresql
  host: localhost
  port: 5432

production:
  database: production_slave_database_name
  <<: *login

master_database:
  database: production_master_database_name
  <<: *login

staging:
  database: staging_database_name
  host: slave-db-pool.local
  <<: *login
  master_database: 
    database: staging_database_name
    host: master-db-server.local
    <<: *login

qa:
  database: qa_master_database_name
  host: qa-master
  <<: *login
  slave_database:
    database: qa_slave_database_name
    host: qa-slave
    <<: *login

development: # Does not use masochism
  database: development_database_name
  <<: *login

If you want a model to always use the Master database, you can inherit ActiveReload::MasterDatabase. Any models with their own database connection will not be affected.

More control at setup

By default, masochism setup! is a shorthand for this:

ActiveReload::ConnectionProxy.setup_for ActiveReload::MasterDatabase, ActiveRecord::Base

The first argument is the model that has the master database connection established; the second argument is the model whose connection gets hijacked by ConnectionProxy. But we don't have to touch ActiveRecord::Base at all:

# set up MyMaster's connection as the master database connection for User:
ActiveReload::ConnectionProxy.setup_for MyMaster, User

The controller filter

If you have any actions you know require the master database for both reads and writes, simply do the following:

# in a controller:
around_filter ActiveReload::MasterFilter, :only => [:show, :edit, :update]
Something went wrong with that request. Please try again.