I’ve been running Kubernetes in my home lab for about a year on Fedora Server instances running on various intel and AMD processors, and about a month ago I experimented with worker nodes running Fedora CoreOS. This has worked very well, and has removed the need to keep all the workers software up to date. I just have to drain, reboot, and then uncordon each node and the whole process is over in about a minute. All good. But when I tried to switch my control plane nodes over to CoreOS, they refuse to create a highly available set of three control nodes. Nodes 2 and 3 fail to join the 1st node. I get one of two errors. 1: “error execution phase preflight: couldn’t validate the identity of the API Server: failed to request the cluster-info ConfigMap: " or 2: " could not find a JWS signature in the cluster-info ConfigMap for token ID”.
I follow the same procedure I use on Fedora Server, where I can set up and break down the cluster with no failures. It’s only when I try to establish a cluster with control nodes on CoreOS that I get the failure. I confirmed that I did not have a firewall in the network. I confirmed that the token the “sudo kubeadm join --control-plane --token=…” asks for is present on the 1st control plane node. The join command is issued right after I do a “sudo kubeadm token create --print-join-command” on the 1st control plane node, so it’s not expired. I can “sudo kubeadm token list” and see the token in the list, but the joining node can’t seem to find it. I’d like to take the next step to get CoreOS on the control plane nodes, but I’m stuck at this point. Any ideas?