janiskemper

makes sense! will do!

> I heard that other clusters (e.g. openstack) have problems at cluster updates where the api is unavailable because the lb adds the nodes before the api server is ready...

@guettli what would be your answer to this question here? > But what do you do if the apiserver becomes unhealthy later and the port is still reachable via tcp?

Your observations are correct @simonostendorf . We do have a check like this in the code that you have the exact behavior you describe. The first node has to be...

closing this. Feel free to re-open!

Closing this due to inactivity. Feel free to re-open

We use the hcloud-go provider for this: https://github.com/syself/cluster-api-provider-hetzner/blob/c6cab3c53bc89782018698a8aec98575a52c3879/pkg/services/hcloud/server/server.go#L379 There is no more logic involved in CAPH to control this. Maybe you can explain a bit more what happens, so that...

@apricote do you know?

Any update here?

amazing, thanks a lot @mrbobbytables :heart_eyes: