CourseAI System Design: The Complete Course · Module 11: Reliability · part 80 of 107
Part 80 · Module 11: Reliability

Lesson 2: Timeouts, retries, breakers and limits under load

34 min read·
Module 11 · Lesson 2 · About 55 minutes · Needs: Module 7 lesson 2 (gateway and fallback chain), Module 2 lesson 3 (residency.py), Module 3 lesson 6 (hard ceilings), Module 8 lesson 4 (idempotency keys). One hosted block of 20 short calls; everything else runs with no key

The situation

Week 11, Monday 11 January 2027. Funmi replays 22 September through Module 7's gateway, with fallbacks now following Module 2 lesson 3's residency.py, which answers "no" for the Gemini API in every region. A timed-out request has nowhere allowed to go: it ends in degraded mode. Gareth asks why the timeout is 8 seconds. Ana asks how many extra calls retries make. Joel asks what stops one tenant's script from slowing everyone.

The running example is trace ask-0616: a Tallis Retail Group request at 13:03:26 UTC whose primary call timed out at 8,000 ms.

The rest of this course is yours to keep

This course is bought on its own, once, and stays readable afterwards, including the parts added to it later.

Or open everything with a plan