Hot Standby vs Replica

Failover patterns.

Overview

Hot standby and read replica are both replicas, but they serve different purposes. Hot standby is for failover: synchronous (or semi-sync) replication, promotable on primary failure with low RPO. Read replica is for read scaling: asynchronous, serves read queries, no promotion path expected. Confusing the two produces architectures that fail at exactly the moment they were supposed to protect.

The approach

Three habits keep the choice rational: hot standby for HA needs, read replica for read scaling, RPO documented per replica so failover behaviour is understood before the failover happens.

Why this compounds

Each correctly-architected replica supports its purpose for years. The team’s database HA fluency deepens; failover stops being a theoretical capability and becomes a tested one.