Checkpoint: Distributed Systems
advanced25 min readLesson 150 of 169
Prove you can reason about redelivery, idempotency, and breakers under an injected clock.
The exam question: a job was delivered, the worker crashed before acknowledging, and the queue delivered it again. Walk through what your system does — in code.
The checkpoint challenge: an idempotency-gated processor over an
at-least-once stream. Jobs arrive twice by design; effects must happen once;
'poison' jobs must exhaust their retry budget into the dead-letter list; and
a failing dependency is recorded, never propagated.