Hi, I explain my problem to you. All my API have two upstream target url for load balancing. Every url point to a server. The question is, if the service in one machine is down, the round robin balancing return always the two url or it execute an health check and remove che upstream url? A clarification, the server where the service is down is not down. Only the service doesn’t work. The health check works only if the server at port X doesn’t responds?
That’s correct - if you have two URLs being checked and have enabled checking them against the round robin then if one starts failing, Tyk will stop sending traffic to one until that test starts to pass again.
Whether the server is up or down doesn’t matter, it’s only if the test starts to report failure.
The list of hosts will remain the same in the API definition, but Tyk will skip the entry until it’s back up.
Which test Tyk run on upstream url? There is a specific error that check by the test to decide if it pass or not?
It’s done either on error (TCP / Downtime) or a non 200 status response