In Tair standard and cluster instances, replica nodes serve a single purpose: high availability (HA) failover. They do not handle read or write requests under normal operation.
| Instance type | Replica node purpose | Handles read/write requests? |
|---|---|---|
| Standard instance (master-replica architecture) | HA failover only | No |
| HA cluster instance (multiple master-replica pairs) | HA failover only | No |
| Read/write splitting instance | Read scaling | Yes — read replicas handle read requests |
When the master node fails, the HA system automatically triggers a master-replica switchover to maintain service continuity.
If your workload is read-heavy, use the read/write splitting architecture. A read/write splitting instance deployed in cloud-native mode supports up to nine read replicas, distributing read traffic across replicas to scale read throughput.
For more information, see Read/write splitting instances.