Redis in Production: Why a Fast Cache Can Still Return Stale Data
Imagine that you have just changed a product’s price. One request is still reading the old price from the database; another has already deleted the old cache entry. Which request writes the cache last? Now imagine thousands of visitors opening the same page just as its cache key expires. How many database queries will you get? Those are the questions I find more interesting than another SET/GET tutorial. This article starts with a small product catalog and follows three problems: stale cache fills, a database overwhelmed by cache misses, and rate limits that behave differently from what their configuration suggests. It ends with eight failures worth simulating before Redis sits on an important request path. ...