Cache invalidation removes or updates cached data when its source changes. It is needed when waiting for the TTL to expire would leave data stale for too long.

• Delete on write: update the database, then delete the related cache key. The next read repopulates it.
• Update on write: update the database and cached value together, as in write-through caching.
• Event-driven invalidation: publish an event when data changes so services or CDN edges can invalidate affected entries.
• Versioned keys: include a version in the key or URL, such as product:42:v3 or app.a1b2c3.js, so new data uses a new cache entry.
• Manual purge: explicitly clear selected keys or paths when content is published or corrected.

Time to Live (TTL)

A TTL defines how long a cached entry remains valid. When the TTL expires, the cache removes the entry or treats it as a miss so fresh data can be fetched from the source.:

• Use a short TTL for frequently changing data.
• Use a long TTL for stable data, such as country lists or versioned static assets.
• Add random variation, or TTL jitter, to prevent many keys from expiring simultaneously and causing a cache stampede.
• A TTL limits staleness but does not guarantee that the cache always contains the latest value.