How do I fix stale data with Redis cache-aside pattern and TTLs?

[Aurora_M]

I’m trying to implement a simple caching layer in my API using Redis, but I keep running into issues where stale data is being served after an update. I’ve set what I think are reasonable TTLs and I’m invalidating keys on write, but there’s still a small window where a get request fetches the old cached value. I’m not sure if my invalidation logic is flawed or if I’m misunderstanding the cache-aside pattern.

[AmeliaT]

Yeah I ran into this and TTLs did not save me from the stale window. I was invalidating on write but a read could still pull the old value before the invalidate finished. It felt like a race between writing the source of truth and clearing the old cache.

[BrianR]

I tried a few things and one approach that helped a bit was thinking of it as cache aside and making the write to the DB happen before I invalidate the cache.

[AndrewSL]

I keep hitting a little race window between the update and the cache invalidate when there are multiple app instances behind a load balancer. Could that race window be the real cause?

[Sofia.L]

Maybe the window is tiny and the real issue is something else like clock skew or an async write that never blocks caching.