Redis was (and is) amazing. In-memory speed, rich data types, pub-sub — a Swiss army knife for backend engineers.
But I learned in 2013 that abusing Redis as a database is a recipe for tears.
What went wrong:
- Used as a cache with no eviction policy — memory spiked fast
- Used for critical state without persistence — lost data during failover
- Used as a queue with no visibility — debugging became guesswork
We fixed it by:
- Using proper write-through patterns
- Adding backups and monitoring keys
- Treating Redis as ephemeral, not durable
Every tool has a sharp edge. You just need to know where it cuts.