node not initialized: failed to find kubelet node IP from cloud provider

[uykusuz]

Hi ccm Team!

I just bootstrapped a single node k8s cluster using hetzner's debian 10 image, did a dist-upgrade and with cilium 1.9.1 as CNI. I followed the instructions on deploy with networks to set it up.

But the node keeps the taint for being uninitialized. CCM complains with the following:

E0118 12:45:34.587422       1 node_controller.go:370] failed to initialize node test at cloudprovider: failed to find kubelet node IP from cloud provider

but on the node the required annotation is present:

alpha.kubernetes.io/provided-node-ip: 10.0.0.6

I tried restarting CCM (perhaps then a refreshed set of node annotations is picked up), that doesn't help.

Do you have any clues what might be happening here?

[uykusuz]

Sorry for the spam. The hcloud secret was wrong and this is how it told me that.

[mway-niels]

I had the same issue. For me the solution was to provide the HCLOUD_NETWORK environment variable.

[ss3rg3]

Same. And setting HCLOUD_NETWORK solved it for me as well. The problem in my case is that I'm starting kubelet with --node-ip to use the private network which the hcloud cloud management controller manager or however that thing is called does not like.

The value is the name of your private network in Hetzner and is stored in the secret hcloud at the key network (see here). You need to add it to the Deployment like this:

- name: HCLOUD_NETWORK
  valueFrom:
    secretKeyRef:
      name: hcloud
      key: network

See ccm-networks.yaml. I used the basic deployment without networks support and it seems to work. Using Cilium as network provider.