Master Redis Cluster Setup: From Basics to Scaling Nodes

Why Use Redis Cluster

Cluster technology is essential for high‑performance website architectures. By sharding data across multiple servers, we reduce the load on any single node and improve response times under high concurrency.

Implementation Strategy

Since Redis 3.0, official support for Redis Cluster is available. Keys are distributed across nodes using hash slots: Redis computes a CRC16 hash of the key, takes the modulo 16384, and maps the resulting slot to a node. This design simplifies adding or removing nodes because slots can be reassigned without restarting the cluster.

Cluster Implementation

Create three directories (6000, 6100, 6200) under a redis_cluster folder, copy redis-server and redis.conf into each, and edit the configuration to set the appropriate port and enable clustering.

daemonize yes
port 6000
cluster-enabled yes
cluster-config-file nodes.conf
cluster-node-timeout 5000
appendonly yes

Start each instance with ./redis-server ./redis.conf, then create the cluster using the Ruby tool:

./redis-trib.rb create --replicas 0 127.0.0.1:6000 127.0.0.1:6100 127.0.0.1:6200

If the redis-trib.rb script is missing, install Ruby and the Redis gem, then retry.

Cluster Changes

To add a new node (e.g., port 6300), create its directory, copy the binaries, adjust redis.conf, and start the instance. Then add it to the cluster:

./redis-trib.rb add-node 127.0.0.1:6300 127.0.0.1:6000

After adding, assign hash slots to the new node using resharding: ./redis-trib.rb reshard 127.0.0.1:6300 Specify the number of slots to move (e.g., 1000) and the destination node ID. The CRC16 algorithm determines which keys map to the newly assigned slots.

How to Add a Slave to Each Master Node

Create a new instance on port 6400, configure it similarly, and add it as a slave:

./redis-trib.rb add-node --slave 127.0.0.1:6400 127.0.0.1:6000

Note that slave nodes accept read commands only; attempts to write (e.g., SET) will return an error.

How to Delete a Node

Attempting to delete a node with data fails until the data is moved. Use resharding to migrate slots away from the node, then delete it:

./redis-trib.rb reshard 127.0.0.1:6000   # move slots to another node
./redis-trib.rb del-node 127.0.0.1:6000 <node-id>

After successful deletion, the node no longer accepts connections.