Worker nodes not visible via kubectl get node on VMs with several network interfaces about kubeadm-ha

[cookeem]

Sorry for reply late, why you use 10.0.2.15:6443 this ip for join? This is your machine ip address not the vm ip address right? I think you should use the vm ip.

kubeadm join --token 7f276c.0741d82a5337f526 10.0.2.15:6443 --discovery-token-ca-cert-hash sha256:c1c15936be9b5c4429cf14074706927a410a150ccb334d6823257cd450f2fe42

[Mar4eli]

I'm not using it, i'm not setting it in create-config.sh, nor putting it anywhere else. It is ip on default network interface eth0 in VM.

When i follow your guide, set VM IP 192.168.120.10 everywhere, kubeadm init cluster with 10.0.2.15:6443. I suppose, iti is becauce eth0 is the default network interface and kubeadm using it if nothing other mentioned. That is why adding advertiseAddress to kubeadm-init.yaml helps.

[cookeem]

Yes, you should join the VM network interface eth1 ip not the machine eth0 ip

[Mar4eli]

So how about adding advertiseAddress to kubeadm-init.yaml.tpl and using K8SHA_IPLOCAL, something like

apiVersion: kubeadm.k8s.io/v1alpha1
kind: MasterConfiguration
kubernetesVersion: v1.9.1
api:
  advertiseAddress: K8SHA_IPLOCAL
networking:
  podSubnet: K8SHA_CIDR
  serviceSubnet: K8SHA_SVC_CIDR
...