Active Record Connection Handler

ConnectionHandler is a collection of ConnectionPool objects. It is used for keeping separate connection pools that connect to different databases.

For example, suppose that you have 5 models, with the following hierarchy:

class Author < ActiveRecord::Base
end

class BankAccount < ActiveRecord::Base
end

class Book < ActiveRecord::Base
  establish_connection :library_db
end

class ScaryBook < Book
end

class GoodBook < Book
end

And a database.yml that looked like this:

development:
  database: my_application
  host: localhost

library_db:
  database: library
  host: some.library.org

Your primary database in the development environment is “my_application” but the Book model connects to a separate database called “library_db” (this can even be a database on a different machine).

Book, ScaryBook, and GoodBook will all use the same connection pool to “library_db” while Author, BankAccount, and any other models you create will use the default connection pool to “my_application”.

The various connection pools are managed by a single instance of ConnectionHandler accessible via ActiveRecord::Base.connection_handler. All Active Record models use this handler to determine the connection pool that they should use.

The ConnectionHandler class is not coupled with the Active models, as it has no knowledge about the model. The model needs to pass a connection specification name to the handler, in order to look up the correct connection pool.

Methods

Class Public methods

new()

📝 Source code
# File activerecord/lib/active_record/connection_adapters/abstract/connection_handler.rb, line 74
      def initialize
        # These caches are keyed by pool_config.connection_name (PoolConfig#connection_name).
        @connection_name_to_pool_manager = Concurrent::Map.new(initial_capacity: 2)
      end
🔎 See on GitHub

Instance Public methods

active_connections?(role = nil)

Returns true if there are any active connections among the connection pools that the ConnectionHandler is managing.

📝 Source code
# File activerecord/lib/active_record/connection_adapters/abstract/connection_handler.rb, line 155
      def active_connections?(role = nil)
        each_connection_pool(role).any?(&:active_connection?)
      end
🔎 See on GitHub

clear_active_connections!(role = nil)

Returns any connections in use by the current thread back to the pool, and also returns connections to the pool cached by threads that are no longer alive.

📝 Source code
# File activerecord/lib/active_record/connection_adapters/abstract/connection_handler.rb, line 162
      def clear_active_connections!(role = nil)
        each_connection_pool(role).each do |pool|
          pool.release_connection
          pool.disable_query_cache!
        end
      end
🔎 See on GitHub

clear_all_connections!(role = nil)

📝 Source code
# File activerecord/lib/active_record/connection_adapters/abstract/connection_handler.rb, line 176
      def clear_all_connections!(role = nil)
        each_connection_pool(role).each(&:disconnect!)
      end
🔎 See on GitHub

clear_reloadable_connections!(role = nil)

Clears the cache which maps classes.

See ConnectionPool#clear_reloadable_connections! for details.

📝 Source code
# File activerecord/lib/active_record/connection_adapters/abstract/connection_handler.rb, line 172
      def clear_reloadable_connections!(role = nil)
        each_connection_pool(role).each(&:clear_reloadable_connections!)
      end
🔎 See on GitHub

connected?(connection_name, role: ActiveRecord::Base.current_role, shard: ActiveRecord::Base.current_shard)

Returns true if a connection that’s accessible to this class has already been opened.

📝 Source code
# File activerecord/lib/active_record/connection_adapters/abstract/connection_handler.rb, line 198
      def connected?(connection_name, role: ActiveRecord::Base.current_role, shard: ActiveRecord::Base.current_shard)
        pool = retrieve_connection_pool(connection_name, role: role, shard: shard)
        pool && pool.connected?
      end
🔎 See on GitHub

connection_pool_list(role = nil)

Returns the pools for a connection handler and given role. If :all is passed, all pools belonging to the connection handler will be returned.

Also aliased as: connection_pools
📝 Source code
# File activerecord/lib/active_record/connection_adapters/abstract/connection_handler.rb, line 93
      def connection_pool_list(role = nil)
        if role.nil? || role == :all
          connection_name_to_pool_manager.values.flat_map { |m| m.pool_configs.map(&:pool) }
        else
          connection_name_to_pool_manager.values.flat_map { |m| m.pool_configs(role).map(&:pool) }
        end
      end
🔎 See on GitHub

connection_pools(role = nil)

establish_connection(config, owner_name: Base, role: Base.current_role, shard: Base.current_shard, clobber: false)

📝 Source code
# File activerecord/lib/active_record/connection_adapters/abstract/connection_handler.rb, line 113
      def establish_connection(config, owner_name: Base, role: Base.current_role, shard: Base.current_shard, clobber: false)
        owner_name = determine_owner_name(owner_name, config)

        pool_config = resolve_pool_config(config, owner_name, role, shard)
        db_config = pool_config.db_config

        pool_manager = set_pool_manager(pool_config.connection_name)

        # If there is an existing pool with the same values as the pool_config
        # don't remove the connection. Connections should only be removed if we are
        # establishing a connection on a class that is already connected to a different
        # configuration.
        existing_pool_config = pool_manager.get_pool_config(role, shard)

        if !clobber && existing_pool_config && existing_pool_config.db_config == db_config
          # Update the pool_config's connection class if it differs. This is used
          # for ensuring that ActiveRecord::Base and the primary_abstract_class use
          # the same pool. Without this granular swapping will not work correctly.
          if owner_name.primary_class? && (existing_pool_config.connection_class != owner_name)
            existing_pool_config.connection_class = owner_name
          end

          existing_pool_config.pool
        else
          disconnect_pool_from_pool_manager(pool_manager, role, shard)
          pool_manager.set_pool_config(role, shard, pool_config)

          payload = {
            connection_name: pool_config.connection_name,
            role: role,
            shard: shard,
            config: db_config.configuration_hash
          }

          ActiveSupport::Notifications.instrumenter.instrument("!connection.active_record", payload) do
            pool_config.pool
          end
        end
      end
🔎 See on GitHub

flush_idle_connections!(role = nil)

Disconnects all currently idle connections.

See ConnectionPool#flush! for details.

📝 Source code
# File activerecord/lib/active_record/connection_adapters/abstract/connection_handler.rb, line 183
      def flush_idle_connections!(role = nil)
        each_connection_pool(role).each(&:flush!)
      end
🔎 See on GitHub

remove_connection_pool(connection_name, role: ActiveRecord::Base.current_role, shard: ActiveRecord::Base.current_shard)

📝 Source code
# File activerecord/lib/active_record/connection_adapters/abstract/connection_handler.rb, line 203
      def remove_connection_pool(connection_name, role: ActiveRecord::Base.current_role, shard: ActiveRecord::Base.current_shard)
        if pool_manager = get_pool_manager(connection_name)
          disconnect_pool_from_pool_manager(pool_manager, role, shard)
        end
      end
🔎 See on GitHub

retrieve_connection_pool(connection_name, role: ActiveRecord::Base.current_role, shard: ActiveRecord::Base.current_shard, strict: false)

Retrieving the connection pool happens a lot, so we cache it in @connection_name_to_pool_manager. This makes retrieving the connection pool O(1) once the process is warm. When a connection is established or removed, we invalidate the cache.

📝 Source code
# File activerecord/lib/active_record/connection_adapters/abstract/connection_handler.rb, line 212
      def retrieve_connection_pool(connection_name, role: ActiveRecord::Base.current_role, shard: ActiveRecord::Base.current_shard, strict: false)
        pool = get_pool_manager(connection_name)&.get_pool_config(role, shard)&.pool

        if strict && !pool
          if shard != ActiveRecord::Base.default_shard
            message = "No connection pool for '#{connection_name}' found for the '#{shard}' shard."
          elsif role != ActiveRecord::Base.default_role
            message = "No connection pool for '#{connection_name}' found for the '#{role}' role."
          else
            message = "No connection pool for '#{connection_name}' found."
          end

          raise ConnectionNotEstablished, message
        end

        pool
      end
🔎 See on GitHub