Actually a great feature. I wish this was documented better. Am I right, that the retries are ignored during the startup and it basically retries the health check, until it succeeds? What happens if the service crashes during startup? Will it be marked as unhealthy?
So the start interval lets you set a separate interval to do health checks during startup. For instance, during startup I might want to run the health check every couple seconds until I know it’s healthy vs once it’s in a stable state, I might only want to check each minute. As to your question about crashing on startup, it’ll be marked as unhealthy once the number of retries have been exceeded
@@JustinHammond ah, okay. So the `retries` parameter is reused? That means setting the retries to a low number and starting to check early and often will most probably lead to an unhealthy container. Doesn't this make docker kill and restart the container? This would be a big error source, if the startup significantly differs on different machines.
@ElCerdoBlanco totally depends. For me, many of my containers start super fast. I use Traefik to route traffic to them but they aren’t discoverable until marked as healthy meaning in my earlier example that a one minute health check means that on deployment there would be one minute of downtime. With this new feature I can set it to check every second during the first 10 seconds so it’s almost seamless. It all depends on how you configure it