PostgreSQL provides read scalability capabilities, owing to Hot Standby and Streaming Replication, two longstanding and popular features.

While this is commonly known, there are a number of issues which become apparent only while performing the actual implementation. For instance:

• Separating reads from writes

• Balancing reads across standby nodes

• The system is only eventually consistent

• Adding extra nodes to a live cluster

• Handling the failure of one node

• Keeping complexity down to a manageable level

In this presentation we will describe the implementation of a read-scalable architecture, from the perspective of how the above issues were tackled, and how successfully.

Finally, we will also consider improvements such as multiple synchronous standby nodes, or the new durability level, introduced in PostgreSQL 9.6.