Switching database connection between readonly one and writable one.

switch_point won't support upcoming ActiveRecord v6.1 or later. Developers should use the builtin multiple database feature introduced in ActiveRecord v6.0. https://guides.rubyonrails.org/active_record_multiple_databases.html Thus the supported ActiveRecord version is v3.2, v4.0, v4.1, v4.2, v5.0, v5.1, and v5.2.

switch_point won't accept any new features. Bug fixes might be accepted. If you'd like to add a new feature (and/or support ActiveRecord >= v6.1), feel free to fork switch_point gem.

1. Upgrade your activerecord gem to v6.0
   • ActiveRecord v6.0 is the only series which supports both builtin multiple database feature and switch_point.
2. Change your application to use ActiveRecord multiple database feature
   • If you'd like to keep the number of connections during this step, it would require some tricks.
3. Remove switch_point gem from your Gemfile
4. Upgrade your activerecord gem to v6.1 or later

Add this line to your application's Gemfile:

gem 'switch_point'

And then execute:

$ bundle

Or install it yourself as:

$ gem install switch_point

Suppose you have 4 databases: db-blog-master, db-blog-slave, db-comment-master and db-comment-slave. Article model and Category model are stored in db-blog-{master,slave} and Comment model is stored in db-comment-{master,slave}.

In database.yml:

production_blog_master:
  adapter: mysql2
  username: blog_writable
  host: db-blog-master
production_blog_slave:
  adapter: mysql2
  username: blog_readonly
  host: db-blog-slave
production_comment_master:
    ...

In initializer:

SwitchPoint.configure do |config|
  config.define_switch_point :blog,
    readonly: :"#{Rails.env}_blog_slave",
    writable: :"#{Rails.env}_blog_master"
  config.define_switch_point :comment,
    readonly: :"#{Rails.env}_comment_slave",
    writable: :"#{Rails.env}_comment_master"
end

In models:

class Article < ActiveRecord::Base
  use_switch_point :blog
end

class Category < ActiveRecord::Base
  use_switch_point :blog
end

class Comment < ActiveRecord::Base
  use_switch_point :comment
end

Article.with_readonly { Article.first } # Read from db-blog-slave
Category.with_readonly { Category.first } # Also read from db-blog-slave
Comment.with_readonly { Comment.first } # Read from db-comment-slave

Article.with_readonly do
  article = Article.first  # Read from db-blog-slave
  article.title = 'new title'
  Article.with_writable do
    article.save!  # Write to db-blog-master
    article.reload  # Read from db-blog-master
    Category.first  # Read from db-blog-master
  end
end

Note that Article and Category shares their connections.

Model.cache and Model.uncached enables/disables query cache for both readonly connection and writable connection.

switch_point also provide a rack middleware SwitchPoint::QueryCache similar to ActiveRecord::QueryCache. It enables query cache for all models using switch_point.

# Replace ActiveRecord::QueryCache with SwitchPoint::QueryCache
config.middleware.swap ActiveRecord::QueryCache, SwitchPoint::QueryCache

# Enable query cache for :nanika1 only.
config.middleware.swap ActiveRecord::QueryCache, SwitchPoint::QueryCache, [:nanika1]

auto_writable is disabled by default.

When auto_writable is enabled, destructive queries is sent to writable connection even in readonly mode. But it does NOT work well on transactions.

Suppose after_save callback is set to User model. When User.create is called, it proceeds as follows.

1. BEGIN TRANSACTION is sent to READONLY connection.
2. switch_point switches the connection to WRITABLE.
3. INSERT statement is sent to WRITABLE connection.
4. switch_point reset the connection to READONLY.
5. after_save callback is called.
   • At this point, the connection is READONLY and in a transaction.
6. COMMIT TRANSACTION is sent to READONLY connection.

Model has several connection-related methods: connection_handler, connection_pool, connected? and so on. Since only connection method is monkey-patched, other connection-related methods doesn't work properly. If you'd like to use those methods, send it to Model.switch_point_proxy.model_for_connection.

There's a proxy which holds two connections: readonly one and writable one. A proxy has a thread-local state indicating the current mode: readonly or writable.

Each ActiveRecord model refers to a proxy. ActiveRecord::Base.connection is hooked and delegated to the referred proxy.

When the writable connection is requested to execute destructive query, the readonly connection clears its query cache.

Basically, each connection managed by a proxy isn't shared between proxies. But there's one exception: ActiveRecord::Base.

If :writable key is omitted (e.g., Nanika1 model in spec/models), it uses ActiveRecord::Base.connection as writable one. When ActiveRecord::Base.connection is requested to execute destructive query, all readonly connections managed by a proxy which uses ActiveRecord::Base.connection as a writable connection clear query cache.

1. Fork it ( https://github.com/eagletmt/switch_point/fork )
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Create a new Pull Request