High Availability

ELI5 - The Vibe Check

High availability means your system keeps running even when a piece of it inevitably fails, because in production, something is always failing somewhere. It's less about preventing failure, which is impossible, and more about making sure a single dead server, a downed data center, or a flaky disk never turns into an outage your users actually notice.

Real Talk

High availability (HA) refers to system design that minimizes downtime by eliminating single points of failure, typically through redundancy across availability zones, automated failover, health checks, and load balancing. It is usually quantified in nines of uptime, such as 99.9% (about 8.7 hours of downtime a year) or 99.99% (about 52 minutes a year), and each additional nine costs disproportionately more engineering effort than the last.

Show Me The Code

# minimal HA setup: 3 instances across 3 zones, auto-failover
replicas: 3
zones: [us-east-1a, us-east-1b, us-east-1c]
healthCheck:
  path: /healthz
  failureThreshold: 3
  autoFailover: true

When You'll Hear This

"We need high availability on the payment service, that one can't go down." / "Three nines sounded fine until we did the math on the downtime budget."