How does Sentinel differ from Redis Cluster?

Redis Sentinel and Redis Cluster solve different problems:

#Redis Sentinel, a monitoring and automatic-failover system

Main goal: to provide high availability for a single logical Redis instance.

How it works:

• Sentinel watches the master and its replicas.
• If the master stops responding, the Sentinels hold a vote and promote a replica to master.
• Clients find out about it and automatically switch to the new master.

What it gives you:

• Automatic failover (switching over on failure).
• Monitoring the nodes' state.
• Reassigning master/replica roles.

What it doesn't have:

• Sentinel doesn't split data across nodes (every copy holds identical data).
• There's no scaling by data volume.

It's used when you need:

A single logical Redis, with several copies for reliability.

#Redis Cluster, distributed data storage

Main goal: to provide both scaling and high availability by splitting data across nodes.

How it works:

• Keys are distributed across several master nodes via hash slots (16384 in total).
• Every master has one or several replicas.
• If a master fails, its replica automatically becomes the new master.

What it gives you:

• Horizontal scaling (much more data can be stored).
• High availability with no Sentinel needed.
• Automatic recovery and request routing.

What it doesn't have:

• No full isolation, some data can become unavailable if a master and its replicas fail at the same time.
• Some commands (multi-key) work in a limited way.

It's used when you need:

A large volume of data, spread across several servers.

#A quick comparison:

Summary:

• Sentinel makes Redis fault-tolerant, but not scalable.
• Cluster makes Redis both fault-tolerant and scalable, by distributing data.