We’re creating a service where availability is really important. Therefore we need our origin servers to be detected unhealthy or unresponsive as soon as possible. We plan to do some production availability testing, by taking down origin servers and pools at random times, to make sure the system failover handles it correctly.
At the moment the documentation of the Cloudflare load balancing is not crystal clear, so I need some clarification.
Am I correct in assuming that the load balancer tries to redirect a connect request to a healthy server, but if that origin server is unresponsive it will detect this and instead send this connect request to another healthy origin server until it reaches a responsive one?
How do I calculate the expected load balancing monitoring health checks sent to a single origin server per minute? My guess right now is, (60 / interval-in-seconds) * number-of-regions. Is this correct?