Since the release of PostgreSQL 9.0, DBAs have had access to asynchronous streaming replication. Unlike the older hot standby methods used in earlier versions, replica servers can connect to an upstream PostgreSQL server and consume data modifications directly. With low network latency and fast transactions, this means that it is fairly common for streaming replicas to lag behind the master by only a few milliseconds.

In the context of high availability, this means we can scale horizontally by copying the database to multiple servers. Of course, this means that we need to copy the entire database to each server. For small-to medium-sized database instances, this is a relatively minor requirement. This also means that we can produce up-to-date backups, perform ad hoc queries on practically live data, and aggregate information into reports without disrupting our primary database.

This recipe will explain how to set up a streaming asynchronous database replica...