Distributing reads to replicas (cluster mode disabled) - Amazon ElastiCache
Services or capabilities described in AWS documentation might vary by Region. To see the differences applicable to the AWS European Sovereign Cloud Region, see the AWS European Sovereign Cloud User Guide.

Distributing reads to replicas (cluster mode disabled)

For cluster mode disabled clusters with replicas, you can use Lettuce's StaticMasterReplicaTopologyProvider with the ReadFrom setting to direct writes to the primary node and distribute reads across replicas. This improves read throughput by offloading reads from the primary.

Use the primary endpoint for writes and the reader endpoint for reads. The reader endpoint automatically load-balances connections across available replicas.

// Primary endpoint for writes, reader endpoint for read distribution
RedisURI primaryURI = RedisURI.Builder.redis("primary-endpoint.cache.amazonaws.com")
    .withPort(6379).withSsl(true).build();
RedisURI readerURI = RedisURI.Builder.redis("reader-endpoint.cache.amazonaws.com")
    .withPort(6379).withSsl(true).build();

RedisClient redisClient = RedisClient.create(clientResources);
redisClient.setOptions(clientOptions);

// Open a primary-replica connection using the StaticMasterReplicaTopologyProvider
StatefulRedisMasterReplicaConnection<String, String> connection =
    MasterReplica.connect(redisClient, StringCodec.UTF8, Arrays.asList(primaryURI, readerURI));

// Direct reads to replicas, writes always go to the primary
connection.setReadFrom(ReadFrom.REPLICA_PREFERRED);

The following ReadFrom options are available:

  • REPLICA_PREFERRED – Reads from replicas when available and falls back to the primary (recommended for most use cases).

  • REPLICA – Reads only from replicas. Fails if no replica is available.

  • PRIMARY_PREFERRED – Reads from the primary, falls back to replicas.

  • ANY_REPLICA – Reads from any replica node.

  • PRIMARY – Reads only from the primary node. Use when strong consistency is required.

Note

Data read from replicas might be slightly stale because replication is asynchronous. If your application requires strong consistency, use ReadFrom.PRIMARY or ReadFrom.PRIMARY_PREFERRED.