Failure modes & availability
Failure modes & availability
Bronze
Components crash, slow down or return errors. Availability is the share of time a system works; dependencies in a chain multiply their availabilities, so every extra hop lowers the total.
- Use when
- Setting availability targets · Deciding where redundancy pays off
- Avoid when
- Treating the math as exact — it assumes independent failures
- Trade-offs
- Every hard dependency added to a request lowers its availability · More copies add nines only if they fail independently
- Requires
- Client–server & request/response
3 objectives · 6 misconceptions
- predict Compute the availability of a request path through several components in a chain.
- explain Explain why a slow dependency can hurt more than a dead one.
- predict Compute how a redundant pair raises availability.
- Not true: Failure means errors; slowness is fine.
- Not true: Adding a 99.9% component keeps the system at 99.9%.
- Not true: Backups are redundancy — the site keeps running while you restore.
- Not true: Every service a system calls lowers its availability the same way.
- Not true: Two copies of a component never fail at the same time.
- Not true: A slow dependency only affects the requests that call it.
Learn
Failure modes and availability~9 minPractice
No katas practice this yet.