High Availability

Introduction

HA is about surviving failures invisibly. Plan for: machine failures, AZ failures, region failures, dependency failures — and the human deploying at 5pm on Friday.

Understanding the topic

• 🪪 Redundancy — N replicas, multi-AZ, multi-region for tier-0.
• ❤️ Health checks — readiness vs liveness done right.
• 🚦 Load balancing — least-connections, sticky only when forced.
• 🛡 Failover — primary/replica DB with automatic promotion.
• 🌀 Graceful shutdown — drain connections, wait for in-flight requests.
• 🔁 Retries with jittered backoff — never thundering-herd.
• 📜 Runbooks — every alert links to one.