Alex Hyett on YouTube · Open on YouTube
Running production: reliability, observability, performance and security · Proficient · Lesson 73 of 98 · 8 min video, then about 6 hours of practice
Resilience: timeouts, retries, idempotency and circuit breakers
After this you can make a service degrade gracefully when its dependencies are slow or failing, without causing retry storms.
Why it matters
Every network call can fail, hang or succeed twice. Resilient services set timeouts on every call, retry only idempotent operations with exponential backoff and jitter, use circuit breakers to stop hammering a failing dependency, and design APIs to be safely retried with idempotency keys. Trap: retries without backoff. They turn a small outage in one service into a large outage everywhere.
How to use this resource
Idempotency explained with how to implement it in an API.