Multidb

A simple, no-nonsense ActiveRecord extension which allows the application to switch between multiple database connections, such as in a master/slave environment. For example:

Multidb.use(:slave) do
  @posts = Post.all
end

The extension was developed in order to support PostgreSQL 9.0's new hot standby support in a production environment.

Randomized balancing of multiple connections within a group is supported. In the future, some kind of automatic balancing of read/write queries could be implemented.

Requirements

Ruby 1.9.3 or later.
ActiveRecord 3.0 or later. (Earlier versions can use the gem version 0.1.10.)

Comparison to other ActiveRecord extensions

Compared to other, more full-featured extensions such as Octopus and Seamless Database Pool:

Minimal amount of monkeypatching magic. The only part of ActiveRecord that is overridden is ActiveRecord::Base#connection.

Non-invasive. Very small amounts of configuration and changes to the client application are required.

Orthogonal. Unlike Octopus, for example, connections follow context:

Multidb.use(:master) do
  @post = Post.find(1)
  Multidb.use(:slave) do
    @post.authors  # This will use the slave
  end
end

Low-overhead. Since connection is called on every single database operation, it needs to be fast. Which it is: Multidb's implementation of connection incurs only a single hash lookup in Thread.current.

However, Multidb also has fewer features. At the moment it will not automatically split reads and writes between database backends.

Getting started

Add to your Gemfile:

gem 'ar-multidb', :require => 'multidb'

All that is needed is to set up your database.yml file:

production:
  adapter: postgresql
  database: myapp_production
  username: ohoh
  password: mymy
  host: db1
  multidb:
    databases:
      slave:
        host: db-slave

Each database entry may be a hash or an array. So this also works:

production:
  adapter: postgresql
  database: myapp_production
  username: ohoh
  password: mymy
  host: db1
  multidb:
    databases:
      slave:
        - host: db-slave1
        - host: db-slave2

If multiple elements are specified, Multidb will use the list to pick a random candidate connection.

The database hashes follow the same format as the top-level adapter configuration. In other words, each database connection may override the adapter, database name, username and so on.

To use the connection, modify your code by wrapping database access logic in blocks:

Multidb.use(:slave) do
  @posts = Post.all
end

To wrap entire controller requests, for example:

class PostsController < ApplicationController
  around_filter :run_using_slave, only: [:index]

  def index
    @posts = Post.all
  end

  def edit
    # Won't be wrapped
  end

  def run_using_slave(&block)
    Multidb.use(:slave, &block)
  end
end

You can also set the current connection for the remainder of the thread's execution:

Multidb.use(:slave)
# Do work
Multidb.use(:master)

Note that the symbol :default will (unless you override it) refer to the default top-level ActiveRecord configuration.

Development mode

In development you will typically want Multidb.use(:slave) to still work, but you probably don't want to run multiple databases on your development box. To make use silently fall back to using the default connection, Multidb can run in fallback mode.

If you are using Rails, this will be automatically enabled in development and test environments. Otherwise, simply set fallback: true in database.yml:

development:
  adapter: postgresql
  database: myapp_development
  username: ohoh
  password: mymy
  host: db1
  multidb:
    fallback: true

Limitations

Multidb does not support per-class connections (eg., calling establish_connection within a class, as opposed to ActiveRecord::Base).

lukaszx0 / multidb