Cache Invalidation Strategies & The Two Hard Things

Computer scientist Phil Karlton famously declared:

"There are only two hard things in Computer Science: cache invalidation and naming things."

When an underlying database record mutates, the corresponding cached representation in Redis, Memcached, or CDN edge nodes becomes stale. If the invalidation mechanism fails, clients will observe outdated data (e.g., viewing an expired product discount, seeing an incorrect bank balance, or viewing a deleted account).

Why Application-Level Invalidation Breaks Down:

1. Uncaught Mutation Paths: A developer runs an emergency SQL script in psql, a nightly ETL batch script updates prices, or a background worker mutates orders. None of these trigger application-level Redis del() calls, leaving stale data in the cache forever.
2. Network Failures on Invalidation: If the application updates the database but network packet loss drops the redis.del() command, the cache remains corrupted until the key's TTL expires.