Slave Groups

A slave group is a group of slaves to which queries are distributed. In other words, all the slaves in the same slave group are load-balanced. Octopus does not load-balance between slaves across multiple slave groups.

Octopus picks slaves in a slave group using a round robin algorithm same as the Multiple Slaves scenario which does not utilize slave groups.

Slaves groups can be used concurrently with Sharding.

Use Case

An usecase of slave groups is to distribute/completely seperate loads of SELECT queries from each part of an application to seperate databases. For example, consider that you have a Web application consists of the A part which serves Web pages for users, and the B part which serves Web pages for admins. If you just send heavy SELECT queries while serving for admins, they will degrade performance serving Web pages for other users. This is where slave groups come in. You can configure seperate slave groups for A and B to make them not affect each other's performance.

Configuration

You can configure any number of slave groups in shards.yml.

Without Sharding

To configure two slave groups slave_group_a and slave_group_b for the development environment, you need shards.yml like:

octopus:
  replicated: true
  environments:
    - development
  development:
    slave_group_a:
      slave1:
        # ...
      slave2:
        # ...
    slave_group_b:
      slave3:
        # ...
      slave4:
        # ...

Then:

• In Octopus.using(slave_group: :slave_group_a) blocks, all the SELECT queries are distributed between slave1 and slave2
• In Octopus.using(slave_group: :slave_group_b) blocks, all the SELECT queries are distribuetd between slave3 and slave4
• Without using, all the SELECT queries go to the development database configured in database.yml

With Sharding

To use slave groups with sharding, you need shards.yml like:

octopus:
  replicated: true
  environments:
    - development
  development:
    shard_1:
      slave_group_a:
        slave1:
          # ...
        slave2:
          # ...
      slave_group_b:
        slave3:
          # ...
        slave4:
          # ...
    shard_2:
      slave_group_a:
        slave5:
          # ...

Then:

• In Octopus.using(shard: :shard_1, slave_group: :slave_group_a) blocks, all the SELECT queries are distributed between slave1 and slave2
• In Octopus.using(shard: :shard_1, slave_group: :slave_group_b) blocks, all the SELECT queries are distribuetd between slave3 and slave4
• Without using, all the SELECT queries go to the development database configured in database.yml