Estimate Route 53 failover time from health check interval, failure threshold, record TTL and resolver propagation.
Route 53 marks an endpoint unhealthy only after the failure threshold is met, so detection takes interval multiplied by threshold — 90 seconds with the standard 30-second interval and a threshold of three. Recovery then waits for caches to expire, which is the record TTL plus an allowance for resolvers and clients that hold entries a little longer, and the two phases add up to the real user-visible outage. DNS failover cannot beat the TTL, so a service with a 300-second TTL has a five-minute floor on recovery no matter how aggressive the health checks are — which is why load balancer or Global Accelerator failover is used when seconds matter.
DNS Failover
failover time = interval × failure threshold + record TTL + propagation allowance.
failover time = interval × failure threshold + record TTL + propagation allowance. Route 53 marks an endpoint unhealthy only after the failure threshold is met, so detection takes interval multiplied by threshold — 90 seconds with the standard 30-second interval and a threshold of three. Recovery then waits for caches to expire, which is the record TTL plus an allowance for resolvers and clients that hold entries a little longer, and the two phases add up to the real user-visible outage.
DNS failover cannot beat the TTL, so a service with a 300-second TTL has a five-minute floor on recovery no matter how aggressive the health checks are — which is why load balancer or Global Accelerator failover is used when seconds matter.
This calculator takes 6 inputs: Health check request interval, Failure threshold, Record TTL, Resolver and client propagation allowance, Recovery time objective, Expected failover events per year. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
Query volume, and therefore cost, scales inversely with TTL, and some resolvers enforce a floor regardless of what you publish. A TTL of 60 seconds is the common compromise; below that you pay for queries you cannot rely on being honoured.
A 10-second interval with a threshold of three detects a fault in 30 seconds instead of 90, at a higher monthly price per health check. Confirm the current fast health check rate with AWS, since the premium can be significant across many endpoints.