Windows domain vs Kind Kubernetes cluster
This story comes from memory too, so i'm a bit short on the details. This was one of the few times I had one of the longest debugging/ fixing sessions ever. Split over three tries, it was a classic "looking in the wrong corner" moment and thereby made it into an article, in the hope that it might help, its a good reminder for me anyways.
I'll make it short:
Problem: Name Resolution problems
So the problem was that "Server A," with the test kind cluster, had a working, "default-ish" networking setup, but "Server B," with exactly the same setup, had completely broken communications. You can imagine the debugging I went through... Fairly quickly, I discovered the problem: any DNS resolution wouldn’t work. Everything was trying to reach the Windows domain servers we had. It turned out that "Server B" was an Ubuntu server that was Windows domain-joined, and the cluster was trying to resolve everything to their DNS servers including the svc.cluster.local addresses, etc.
I don't know why it was domain joined but i did know the quickest solution:
Remove from the domain
I replaced systemd-resolved with dnsmasq, as i didn't care at this point. That seemed to have worked, while keeping the Windows Domain stuff in business.
Proost,