When rails app works with multiple databases, the database performance could be an issue due to the join features is blocked across databases. Even though dblink provides a way for the database to connect an external database, the latency is high if join via dblink. A better solution is to create a materialized view via dblink and cache the data.

The background: One of my rails app, beside the default primary database, it has another PG database handled by the gem secondbase, the second DB contains all the usage stats, which is not important but query intensive, therefore it causing a lot of performance issue, when query stats data.

If there is only one database, and you want to use views to cache SQL Query, the gem scenic does a good job.

Open external connection via dblink

  • Get the dblink connnection str into a initializer
hash = ActiveRecord::Base.connection_config
dblink_hash = {}
dblink_hash[:dbname] = hash[:database]
dblink_hash[:host] = hash[:host] if hash[:host]
dblink_hash[:user] = hash[:username] if hash[:username]
dblink_hash[:password] = hash[:password] if hash[:password]
DBLINK_INFO = { |k, v| "#{k}=#{v}" }.join(' ')

connection str options

  • SQL via dblink
SELECT * FROM dblink('#{DBLINK_INFO}', 'SELECT STR against the external db')

Create Materialized View in migration

Create a Materialized view in rails migration

      SELECT * FROM dblink('#{DBLINK_INFO}',
        'SELECT, FROM users')
      AS second_users(user_id uuid, name text)

this will create a second_users in the secondbase, which will be a copy of the users tables

Syncing Materialized View

Since PostgreSQL deals with the Mview the same way as table, so we can create a model against the view.

class SecondUser < SecondBase::Base # ActiveRecord::Base
  has_many :associations, foreign_key: :user_id

  def self.refresh_materialized_view(concurrently = false)
    # postgresql_version >= 9.4 if concurrently = true
    connection.execute "REFRESH MATERIALIZED VIEW #{concurrently ? 'CONCURRENTLY' : ''} second_users;"

Finally set a rake task to sync daily;