How to Stop Redis Cache Penetration, Avalanche, and Breakdown

Key Concepts

Cache penetration occurs when many requests query keys that do not exist; cache avalanche happens when a large number of keys expire simultaneously; cache breakdown (or stampede) occurs when a hot key expires while many users request it.

Solutions for Cache Penetration

Cache null values (e.g., store key with value null and a short TTL).

Set whitelist/blacklist to block malicious IDs.

Use a Bloom filter (or BitMap) to pre‑filter valid keys before hitting Redis.

Apply network‑level protection such as firewalls.

Solutions for Cache Avalanche

Stagger key expiration times by adding random offsets to TTL.

Monitor hot data and adjust TTL in real time.

Use lock mechanisms to serialize DB fallback when a cache miss occurs.

Employ multi‑level caching (NGINX cache, Redis, other caches) for higher reliability.

Mark cache status and refresh expired keys asynchronously.

Solutions for Cache Breakdown

Pre‑warm hot data in cache before it becomes popular.

Continuously monitor hot keys and dynamically adjust their TTL.

Use locks to prevent a thundering‑herd effect when the hot key is missing.

The root cause of all three scenarios is a drop in Redis hit rate, causing requests to fall back to the database, which can quickly become overloaded.

Additional precautions: set appropriate TTL for null entries, combine blacklist with firewall rules, and be aware of Bloom filter false‑positive possibilities.