High availability (HA)
High availability (HA) is the property of a system that keeps working for long periods despite component failures. It is achieved by removing single points of failure: running redundant copies, spreading them across failure domains such as availability zones, and failing over automatically when one breaks. Availability is often expressed in "nines", such as 99.99%.
How it works
HA designs combine redundancy, health checks and automatic failover. A load balancer stops sending traffic to an unhealthy instance, and a database promotes a replica when the primary dies.
The math rewards independent redundancy. If one instance is available 99% of the time, two independent instances with failover give 1 minus 0.01 x 0.01 = 99.99%, if failures are truly independent. Each nine matters: over a year, 99.9% allows about 8.8 hours of downtime and 99.99% about 52.6 minutes.
Why it matters in production
HA is how teams meet a service level objective without heroics. Shared dependencies break the math, so three replicas in one zone are not HA. Test failover with chaos engineering, because untested failover often fails.
See also
On the One2N blog
Deploying a scalable NATS cluster part 1: core architecture and considerations
Gotchas to avoid when using AWS spot instances in production
Further reading
