使用 Redis Sentinel 实现高可用

When using Redis replication, what happens if your master server suddenly fails? Your application might go down!

This is where Redis Sentinel comes in. It's designed to make your Redis deployments highly available by automatically handling master failures.

Redis Sentinel isn't just one process; it's a distributed system of Sentinel instances. Each Sentinel has three main jobs:

• Monitoring: Constantly checks if your master and replica instances are working correctly.
• Notification: Alerts system administrators or other computer programs when a monitored Redis instance goes wrong.
• Automatic Failover: If a master fails, Sentinel can start a failover process where a replica is promoted to master, and other replicas are reconfigured.

A typical Sentinel setup involves:

• One Redis master instance.
• One or more Redis replica instances.
• At least three Sentinel instances running on separate servers.

The Sentinels constantly communicate with each other and with the Redis instances to maintain a consistent view of the system's health.

Each Sentinel instance needs a configuration file, typically named sentinel.conf. Here's a basic example. The key directive is sentinel monitor:

• mymaster: The name you give your master set.
• 127.0.0.1 6379: The master's IP and port.
• 2: The quorum, explained next.

The quorum (the 2 in our config) is a crucial concept. It's the minimum number of Sentinels that must agree that a master is down before a failover process can be initiated.

This prevents a single Sentinel (or a network issue affecting only one Sentinel) from triggering an unnecessary failover, ensuring reliability.

Once your sentinel.conf file is ready, you can start a Sentinel instance using the redis-sentinel command, pointing it to your config file.

You would typically run multiple Sentinels on different machines for true high availability.

You can connect to a Sentinel instance using redis-cli (using the Sentinel's port, e.g., 26379) and query its status. The info sentinel command provides a wealth of information.

When a master fails:

1. Sentinels detect the master is unreachable.
2. If enough Sentinels (the quorum) agree, they elect a leader Sentinel.
3. The leader Sentinel promotes one of the healthy replicas to be the new master.
4. The other replicas are reconfigured to replicate from the new master.
5. Applications are notified of the new master's address.

Applications don't connect directly to the Redis master's IP and port. Instead, they are configured to connect to one or more Sentinel instances.

The client library then asks Sentinel for the current master's address. This means applications don't need to be reconfigured manually after a failover; Sentinel handles it transparently!

Which of the following is NOT a primary function of Redis Sentinel?

You've learned that Redis Sentinel is essential for creating highly available Redis deployments. It monitors your instances, performs automatic failovers, and keeps your applications connected even when a master goes down.

By using Sentinel, you significantly reduce downtime and improve the resilience of your Redis-backed applications. Next, we'll explore Redis Cluster for even greater scalability!