GuideIntermediate

Resilience And Reliability Patterns

Learn to make a distributed system survive the failure of its parts: timeouts, retries with backoff and jitter, idempotency, circuit breakers, bulkheads, graceful degradation, and load shedding. The guide's core is partial failure — in a distributed system something is always down or slow — and how to stop a dependency's failure from taking down the whole system. You work through a single case study from start to finish: Mercado's checkout flow, where `orders` calls `payments` and `shipping`, and those dependencies genuinely fail and slow down. Everything is simulated and measured in Python with configurable failure rates and latency: you'll measure thread exhaustion without a timeout, a retry storm without backoff, how many calls a circuit breaker saves when a service is dead, how a bulkhead isolates a slow dependency so it doesn't drain the others, and the checkout's success rate before and after hardening it. The capstone combines every pattern over Mercado's full checkout flow and measures the result.