Cache Invalidation

* Scheduling Guide

Cache invalidation jobs need to run before users encounter stale data, which usually means early morning or at the start of a business cycle. The timing depends on how the cache is populated: if content changes on a predictable schedule (nightly ETL, daily CMS publish), run the invalidation job a few minutes after the upstream process completes. For CDN purges, batch invalidation is far more efficient than per-asset requests — run a single scheduled sweep rather than triggering on each change. Choose intervals based on your cache TTL: a one-hour TTL rarely needs a sub-hourly invalidation job, and over-invalidating can cause cache stampedes that defeat the purpose of caching entirely.