Redis 영속성과 HA

Imagine your Redis server suddenly crashes! Without a plan, all the valuable data stored in its memory would be lost forever. This is where persistence comes in.

Redis persistence means saving your in-memory dataset to disk. This ensures that even if Redis restarts or crashes, your data can be restored from the saved copy.

Redis Database (RDB) persistence works by taking periodic snapshots of your dataset. It creates a compact, point-in-time representation of your data.

• Pros: Very fast for backups and restores, compact file size.
• Cons: Potential for data loss between snapshots if a crash occurs.

You can configure RDB snapshots in your redis.conf file. The save directive tells Redis when to take a snapshot. Here's an example:

Append Only File (AOF) persistence records every write operation received by the server. Instead of snapshots, it's a log of commands that can be replayed to reconstruct the dataset.

• Pros: Much less data loss risk (can be configured for every write), human-readable log.
• Cons: Larger file size, potentially slower restore than RDB.

AOF persistence is enabled with appendonly yes. You can control how often Redis syncs the AOF file to disk using appendfsync:

Both RDB and AOF have their strengths. Often, the best approach is to use both:

• AOF: For maximum data safety, as it logs almost every write.
• RDB: For fast backups and disaster recovery, especially for cold starts.

If both are enabled, Redis will use the AOF file to reconstruct the dataset upon restart, as it generally guarantees less data loss.

Beyond persistence, High Availability (HA) ensures your Redis service remains operational even if a server fails. This is crucial for critical applications that cannot tolerate downtime.

Redis offers two main strategies for HA: Sentinel and Cluster. Both involve multiple Redis instances working together to provide redundancy and automatic failover.

Redis Sentinel is a system designed to help manage Redis instances. It provides:

• Monitoring: Constantly checks if your master and replica instances are working correctly.
• Notification: Alerts system administrators when something goes wrong.
• Automatic Failover: If a master instance fails, Sentinel promotes a replica to master and reconfigures others.
• Configuration Provider: Clients connect to Sentinels to discover the current master.

Redis Cluster provides a way to run Redis across multiple nodes, offering both sharding (distributing data) and high availability.

It automatically splits your dataset among multiple Redis instances (nodes). If a master node fails, its replica is automatically promoted. This allows for horizontal scaling and ensures your data is always accessible, even with node failures.

In this lesson, we explored how to protect your Redis data and ensure continuous service. We covered:

• RDB persistence: Snapshotting your data at intervals for fast backups.
• AOF persistence: Logging every write operation for maximum data durability.
• Redis Sentinel: Monitoring and automatic failover for single-instance setups.
• Redis Cluster: Distributing data across nodes for scalable high availability.

Understanding these concepts is vital for building robust, production-ready Redis applications!