🔒 Paid

Cache Revalidation Herd

A batch of your most-requested catalog objects was warmed into the shared cache at the same time and shares a single, unjittered expiry. That shared TTL has just been crossed: cache misses on these objects are appearing across the entire API fleet almost simultaneously, and origin is seeing a volume of refresh traffic far beyond anything its steady-state load would suggest. These are read-heavy, non-transactional objects — brief staleness after a refresh is tolerable, but origin has a hard operating ceiling and the catalog service behind it also serves other, latency-sensitive traffic you cannot afford to degrade. Diagnose why origin load has spiked so far past normal and bring it back to a state that will hold the next time this batch of objects expires, not just this one.

Difficulty
Senior
Expected Time
~15 min
Category
API & Traffic

Skills

  • Diagnosing synchronized server-side cache expiry
  • Request coalescing / single-flight refresh design
  • TTL jitter as a preventative fix vs. a reactive one
  • Recognizing when added capacity multiplies duplicate work instead of absorbing it

System Architecture

  • Clients
  • API Fleet
  • Shared Cache
  • Origin / Catalog Service

This is a Pro scenario. Upgrade to Pro to unlock it.

Upgrade to ProBack to Incidents