ActiveRecord::Turntable

ActiveRecord::Turntable is a database sharding extension for ActiveRecord.

Dependencies

activerecord(>=5.0.0, <6.0)

• Current latest stable branch is stable branch.

If you are using with older activerecord versions, use following versions.

• activerecord 4.x - use activerecord-turntable version 2.x.(stable-2-x branch)
• activerecord 3.x - use activerecord-turntable version 1.x.(stable-1-x branch)

Supported DBMS

MySQL only.

Installation

Add to Gemfile:

gem 'activerecord-turntable', '~> 4.4.1'

Run a bundle install:

bundle install

Run install generator:

bundle exec rails g active_record:turntable:install

generator creates #{Rails.root}/config/turntable.yml

Terminologies

Shard

Shard is a database which is horizontal partitioned.

Cluster

Cluster of shards. i.e) set of userdb1, userdb2, userdb3. Shards in the same cluster should have the same schema structure.

Master

Default ActiveRecord::Base's connection.

Sequencer

Turntable's sequence system for clustered database.

This keeps primary key ids to be unique each shards.

Example

Example Databases Structure

One main database(default ActiveRecord::Base connection) and three user databases sharded by user_id.

                  +-------+
                  |  App  |
                  +-------+
                      |
       +---------+---------+---------+---------+
       |         |         |         |         |
  `--------` `-------` `-------` `-------` `-------`
  | Master | |UserDB1| |UserDB2| |UserDB3| | SeqDB |
  `--------` `-------` `-------` `-------` `-------`

Example Configuration

Edit DSL type configuration file(config/turntable.rb) or YAML configuration file(turntable.yml). And add database connection settings to database.yml.

See below example config.

• example turntable.rb

cluster :user_cluster do
  # algorithm [algorithm name symbol(range|range_bsearch|modulo)]
  algorithm :range_bsearch

  # sequencer [sequence name] [sequence type] [*options hash]
  sequencer :user_seq, :mysql, connection: :user_seq

  # shard [range], to: [connection names in database.yml]
  shard   1...100,        to: :user_shard_1
  shard 100...200,        to: :user_shard_2
  shard 200...2000000000, to: :user_shard_3

  # If you are using modulo algorithm, pass integer sequence start with zero.
  #
  # shard 0, to: :user_shard_1
  # shard 1, to: :user_shard_2
  # shard 2, to: :user_shard_3
end

• example turntable.yml

    development:
      clusters:
        user_cluster: # <-- cluster name
          algorithm: range_bsearch # <-- `range`, `range_bsearch` or `modulo`
          seq:
            user_seq: # <-- sequencer name
              seq_type: mysql # <-- sequencer type
              connection: user_seq_1 # <-- sequencer database connection setting
          shards:
            - connection: user_shard_1 # <-- shard name
              less_than: 100           # <-- shard range(like mysql partitioning) If you are using a modulo algorithm, it doesn't need it.
            - connection: user_shard_2
              less_than: 200
            - connection: user_shard_3
              less_than: 2000000000

• database.yml

    connection_spec: &spec
      adapter: mysql2
      encoding: utf8
      reconnect: false
      pool: 5
      username: root
      password: root
      socket: /tmp/mysql.sock

    development:
      <<: *spec
      database: sample_app_development
      seq: # <-- sequence database definition
        user_seq_1:
          <<: *spec
          database: sample_app_user_seq_development
      shards: # <-- shards definition
        user_shard_1:
          <<: *spec
          database: sample_app_user1_development
        user_shard_2:
          <<: *spec
          database: sample_app_user2_development
        user_shard_3:
          <<: *spec
          database: sample_app_user3_development

About algorithms

• range, range_bsearch - distribute records by key range

  algorithm :range_bsearch

  shard      1...20_000,     to: :user_shard_1
  shard 20_000...40_000,     to: :user_shard_2
  shard 40_000...60_000,     to: :user_shard_1
  shard 60_000...80_000,     to: :user_shard_2
  shard 80_000...10_000_000, to: :user_shard_3

• modulo - distribute records by key modulo

  algorithm :modulo

  shard 0, to: :user_shard_1
  shard 1, to: :user_shard_2
  shard 2, to: :user_shard_3

• hash_slot - distribute records by key hashes

default hash function is Zlib.crc32(key.to_s)

  algorithm :hash_slot
  # Or specify hash function
  # algorithm :hash_slot, hash_func: ->(key) { Zlib.adler32(key.to_s) }

  shard     0...4096,  to: :user_shard_1
  shard  4096...8192,  to: :user_shard_2
  shard  8192...12288, to: :user_shard_3
  shard 12288...16384, to: :user_shard_4

Slave support (experimental)

Slave enabled configuration examples:

• config/turntable.rb

cluster :user_cluster do
...

  # shard [range], to: [connection names in database.yml]
  shard   1...100,        to: :user_shard_1, slaves: [:user_shard_1_1]
  shard 100...200,        to: :user_shard_2, slaves: [:user_shard_2_1]
  shard 200...2000000000, to: :user_shard_3, slaves: [:user_shard_3_1]
end

• config/turntable.yml

    development:
      clusters:
        user_cluster: # <-- cluster name
          ...
          shards:
            - connection: user_shard_1
              less_than: 100
              slaves:
                - user_shard_1_1
            - connection: user_shard_2
              less_than: 200
              slaves:
                - user_shard_2_1
            - connection: user_shard_3
              less_than: 2000000000
              slaves:
                - user_shard_3_1

• config/database.yml

Add slave connection settings under shards.

...
  shards:
    user_shard_1:
      <<: *default
      database: turntable_user_shard_1_test
    user_shard_1_1:
      <<: *default
      database: turntable_user_shard_1_1_test
    user_shard_2:
      <<: *default
      database: turntable_user_shard_2_test
    user_shard_2_1:
      <<: *default
      database: turntable_user_shard_2_1_test
    user_shard_3:
      <<: *default
      database: turntable_user_shard_3_test
    user_shard_3_1:
      <<: *default
      database: turntable_user_shard_3_1_test

Slave usage:

User.with_slave {
  # `User` model will use slave databases within this block.
}

User.with_master {
  # `User` model will use master database within this block.
}

Example Migration

Generate a model:

bundle exec rails g model user name:string

And Edit migration file:

class CreateUsers < ActiveRecord::Migration
  # Specify cluster executes migration if you need.
  # Default, migration would be executed to all databases.
  # clusters :user_cluster

  def change
    create_table :users do |t|
      t.string :name
      t.timestamps
    end
    create_sequence_for(:users) # <-- create sequence table
  end
end

Then please execute rake tasks:

bundle exec rake db:create
bundle exec rake db:migrate

Those rake tasks would be executed to shards too.

Example Model

Add turntable [shard_key_name] to the model class:

class User < ApplicationRecord
  turntable :user_cluster, :id
  sequencer :user_seq
  has_one :status
end

class Status < ApplicationRecord
  turntable :user_cluster, :user_id
  sequencer :user_seq
  belongs_to :user
end

Usage

Creating

    > User.create(name: "hoge")
      (0.0ms) [Shard: user_seq_1] BEGIN
      (0.3ms) [Shard: user_seq_1] UPDATE `users_id_seq` SET id=LAST_INSERT_ID(id+1)
      (0.8ms) [Shard: user_seq_1] COMMIT
      (0.1ms) [Shard: user_seq_1] SELECT LAST_INSERT_ID()
      (0.1ms) [Shard: user_shard_1] BEGIN
    [ActiveRecord::Turntable] Sending method: insert, sql: #<Arel::InsertManager:0x007f8503685b48>, shards: ["user_shard_1"]
      SQL (0.8ms) [Shard: user_shard_1] INSERT INTO `users` (`created_at`, `id`, `name`, `updated_at`) VALUES ('2012-04-10 03:59:42', 2, 'hoge', '2012-04-10 03:59:42')
      (0.4ms) [Shard: user_shard_1] COMMIT
    => #<User id: 2, name: "hoge", created_at: "2012-04-10 03:59:42", updated_at: "2012-04-10 03:59:42">

Retrieving

    > user = User.find(2)
    [ActiveRecord::Turntable] Sending method: select_all, sql: #<Arel::SelectManager:0x007f850466e668>, shards: ["user_shard_1"]
      User Load (0.3ms) [Shard: user_shard_1] SELECT `users`.* FROM `users` WHERE `users`.`id` = 2 LIMIT 1
    => #<User id: 2, name: "hoge", created_at: "2012-04-10 03:59:42", updated_at: "2012-04-10 03:59:42">

Updating

    > user.update_attributes(name: "hogefoo")
      (0.1ms) [Shard: user_shard_1] BEGIN
    [ActiveRecord::Turntable] Sending method: update, sql: UPDATE `users` SET `name` = 'hogefoo', `updated_at` = '2012-04-10 04:07:52' WHERE `users`.`id` = 2, shards: ["user_shard_1"]
      (0.3ms) [Shard: user_shard_1] UPDATE `users` SET `name` = 'hogefoo', `updated_at` = '2012-04-10 04:07:52' WHERE `users`.`id` = 2
      (0.8ms) [Shard: user_shard_1] COMMIT
    => true

Delete

    > user.destroy
      (0.2ms) [Shard: user_shard_1] BEGIN
    [ActiveRecord::Turntable] Sending method: delete, sql: #<Arel::DeleteManager:0x007f8503677ea8>, shards: ["user_shard_1"]
      SQL (0.3ms) [Shard: user_shard_1] DELETE FROM `users` WHERE `users`.`id` = 2
      (1.7ms) [Shard: user_shard_1] COMMIT
    => #<User id: 2, name: "hogefoo", created_at: "2012-04-10 03:59:42", updated_at: "2012-04-10 04:07:52">

Counting

    > User.count
    [ActiveRecord::Turntable] Sending method: select_value, sql: #<Arel::SelectManager:0x007f9e82ccebb0>, shards: ["user_shard_1", "user_shard_2", "user_shard_3"]
       (0.8ms) [Shard: user_shard_1] SELECT COUNT(*) FROM `users`
       (0.3ms) [Shard: user_shard_2] SELECT COUNT(*) FROM `users`
       (0.2ms) [Shard: user_shard_3] SELECT COUNT(*) FROM `users`
    => 1

Sequencer

Sequencer provides generating global IDs.

Turntable has follow 3 sequencers currently:

• :mysql - Use database table to generate ids.
• :barrage - Use barrage gem to generate ids
• :katsubushi - katsubushi sequencer backend

Mysql example

First, add configuration to turntable.yml and database.yml

• database.yml

    development:
      ...
      seq: # <-- sequence database definition
        user_seq_1:
          <<: *spec
          database: sample_app_user_seq_development

• turntable.yml

    development:
      clusters:
        user_cluster: # <-- cluster name
          ....
          seq:
            user_seq: # <-- sequencer name
              seq_type: mysql # <-- sequencer type
              connection: user_seq_1 # <-- sequencer database connection

Add below to the migration:

create_sequence_for(:users) # <-- this line creates sequence table named `users_id_seq`

Next, add sequencer definition to the model:

  class User < ApplicationRecord
    turntable :id
    sequencer :user_seq # <-- this line enables sequencer module
    has_one :status
  end

Barrage example

First, add barrage gem to your Gemfile:

gem 'barrage'

Then, add configuration to turntable.yml:

• turntable.yml

    development:
      clusters:
        user_cluster: # <-- cluster name
          ....
          seq:
            barrage_seq: # <-- sequencer name
              seq_type: barrage # <-- sequencer type
              options: # <-- options passed to barrage
                generators:
                  - name: msec
                    length: 39 # MAX 17.4 years from start_at
                    start_at: 1396278000000 # 2014/04/01 00:00:00 JST
                  - name: redis_worker_id
                    length: 16
                    ttl: 300
                    redis:
                      host: '127.0.0.1'
                  - name: sequence
                    length: 9

Next, add sequencer definition to the model:

  class User < ApplicationRecord
    turntable :id
    sequencer :barrage_seq # <-- this line enables sequencer module
    has_one :status
  end

Katsubushi example

katsubushi is available as a sequence server. ActiveRecord::Turntable accesses katsubushi via Dalli.

First, add dalli gem to your Gemfile:

gem 'dalli'

Then, add configuration to turntable.yml:

• turntable.yml

    development:
      clusters:
        user_cluster: # <-- cluster name
          ....
          seq:
            katsubushi_seq: # <-- sequencer name
              seq_type: katsubushi # <-- sequencer type
              options: # <-- options passed to dalli
                servers:
                  - host: localhost
                    port: 11212
                  - host: localhost
                    port: 11213

Next, add sequencer definition to the model:

  class User < ApplicationRecord
    turntable :id
    sequencer :katsubushi_seq # <-- this line enables sequencer module
    has_one :status
  end

Transactions

Turntable has some transaction support methods.

shards_transaction

Pass AR::Base instances, shards_transaction method suitable shards

user = User.find(2)
user3 = User.create(name: "hoge3")

User.shards_transaction([user, user3]) do
  user.name  = "hogehoge"
  user3.name = "hogehoge3"
  user.save!
  user3.save!
end

cluster_transaction

When executing transaction on all shards in the cluster, use #{cluster_name}_transaction method:

User.user_cluster_transaction do
  # Transaction is opened all shards in "user_cluster"
end

Migration

If you specify cluster or shard, migration will be executed to the cluster(or shard) and master database.

Default, migrations will be executed to all databases.

to specify cluster:

    class CreateUsers < ActiveRecord::Migration
      clusters :user_cluster
      ....
    end

to specify shard:

    class CreateUsers < ActiveRecord::Migration
      shards :user_shard_01
      ....
    end

Limitations

• Queries includes "ORDER BY", "GROUP BY" and "LIMIT" clauses cannot be distributed.
• "has many through" and "habtm" relationships may returns unexpected results. ex) User-Friend-User relation

TIPS

Send query to a specific shard.

Use with_shard method:

    AR::Base.with_shard(shard1) do
      # something queries to shard1
    end

with_shard method accepts following types to specify a shard:

• ActiveRecord::Turntable::Shard object
• AcitveRecord::Base object - AR::Base#turntable_shard will be used
• Numeric, String - a shard key value
• Symbol - shard name symbol

To access shard objects, use below:

• AR::Base.connection.shards # \{shard_name => shard_obj,....}
• AR::Base#turntable_shard # Returns current object's shard
• AR::Base.connection.shard_for(shard_key_value) #=> shard

Send query to all shards

Use with_all method:

  User.with_all do
    User.order("created_at DESC").limit(3).all
  end # => Returns Array of results

Cannot specify error is raised between associations

Normally, activerecord-turntable detects shard keys on associated models, but auto-detection will fail with following conditions:

• foreign key column != shard key column
• Using different shard key names

For example:

class User
  # shard key:
  # foreign key: main_user_item_id
  belongs_to :main_user_item, class_name: "UserItem", required: false
end

class UserItem
  # shard key name: :user_id
  turntable :user_cluster, :user_id
end

This example raises CannotSpecifyShardError

> user.main_user_item
  User Load [Shard: master] (0.4ms)  SELECT  `users`.* FROM `users` ORDER BY `users`.`id` ASC LIMIT 1
[ActiveRecord::Turntable] Error on Building Fader: SELECT  `user_items`.* FROM `user_items` WHERE `user_items`.`id` = 2198059200000 LIMIT 1, on_method: select_all, err: cannot specifyshard for query: SELECT "user_items".* FROM `user_items` WHERE (`user_items`.`id` = 2198059200000) LIMIT 1
ActiveRecord::Turntable::CannotSpecifyShardError: cannot specify shard for query: SELECT "user_items".* FROM `user_items` WHERE (`user_items`.`id` = 2198059200000) LIMIT 1

Use foreign_shard_key option to pass a shard key condition:

-belongs_to :main_user_item, class_name: "UserItem"
+belongs_to :main_user_item, class_name: "UserItem", foreign_shard_key: :id

> user.main_user_item
  User Load [Shard: master] (0.2ms)  SELECT  `users`.* FROM `users` ORDER BY `users`.`id` ASC LIMIT 1
[ActiveRecord::Turntable] Sending method: select_all, sql: #<Arel::SelectManager:0x007f8080bd0670>, shards: ["user_shard_1"]
Changing UserItem's shard to user_shard_1
  UserItem Load [Shard: user_shard_1] (0.2ms)  SELECT  `user_items`.* FROM `user_items` WHERE `user_items`.`user_id` = 1 AND `user_items`.`id` = 2198059200000 LIMIT 1
Changing UserItem's shard to master
=> #<UserItem id: 2198059200000, user_id: 1, item_id: 1, created_at: "2017-05-23 04:41:13", updated_at: "2017-05-23 04:41:13">

Performance Exception

To notice queries causing performance problem, Turntable has follow options.

• raise_on_not_specified_shard_query - raises on queries execute on all shards
• raise_on_not_specified_shard_update - raises on updates executed on all shards

Add to turntable.yml or turntable.rb:

development:
   ....
   raise_on_not_specified_shard_query: true
   raise_on_not_specified_shard_update: true

# Write on top level
raise_on_not_specified_shard_query true
raise_on_not_specified_shard_update true

Thanks

ConnectionProxy, Distributed Migration implementation is inspired by Octopus and DataFabric.

License

activerecord-turntable is released under the MIT license:

Copyright (c) 2012 Drecom Co.,Ltd.

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.