Rke: Add nodelocal DNS option

[superseb]

Next to kubedns and CoreDNS itself, support setting nodelocal option as released in https://github.com/kubernetes/kubernetes/pull/70555 for v1.13

[vikas027]

Hey @superseb ,

Is there a workaround by tweaking a file or configMap for the time being?

My use case is that we use dnsmasq on the host as well to resolve internal domains by pointing to specific nameservers. In order to do that, I need to have IP 172.17.0.1 (default) in /etc/resolv.conf of the docker container.

[deniseschannon]

We should add this as an option in k8s 1.15-k8s 1.18 templates for when we release 2.4 but not turned on by default.

[deniseschannon]

Available with v1.1.0-rc10

[superseb]

Can be tested using:

dns:
  provider: coredns
  nodelocal:
    ip_address: "169.254.20.10"

Things to test:

• --cluster-dns on the kubelet container should be set to this address
• internal and external dns resolution should work

[jiaqiluo]

the enhancement is validated with RKE version v1.1.0-rc10

>./rke -v
rke version v1.1.0-rc10

Use RKE to provision a cluster with the following yml file

nodes:
  - address: 
    internal_address: 
    user: ubuntu
    role: [etcd, controlplane, worker]
    ssh_key_path:  
  - address: 
    internal_address: 
    user: ubuntu
    role: [etcd, controlplane, worker]
    ssh_key_path:  
  - address:  
    internal_address: 
    user: ubuntu
    role: [etcd, controlplane, worker]
    ssh_key_path:  

dns:
  provider: coredns
  nodelocal:
    ip_address: "169.254.20.10"

Do the following checks on the cluster:

• confirm that a daemonset named node-local-dns is deployed in the cluster

> k get daemonsets.apps -n kube-system
NAME             DESIRED   CURRENT   READY   UP-TO-DATE   AVAILABLE   NODE SELECTOR   AGE
canal            3         3         3       3            3           <none>          20m
node-local-dns   3         3         3       3            3           <none>          20m

• confirm the --cluster-dns on the kubelet container should be set to designated address (need to ssh into nodes)

ubuntu@ip-172-31-25-27:~$ docker inspect kubelet | grep -e  'cluster-dns'
            "--cluster-dns=169.254.20.10",
                "--cluster-dns=169.254.20.10",

• confirm that the internal and external dns resolution works following the checsk on https://rancher.com/docs/rancher/v2.x/en/troubleshooting/dns/

[jiaqiluo]

This is also validated in rancher:master-head 8ede17d10

k8s version:

• v1.17.3-rancher1-2
• v1.16.7-rancher1-2
• v1.15.10-rancher1-2