Cattle crashes by missing docker socket when deploying Rancher UI

[DMW007]

Describe the bug
When installing rancher, the UI works but show different errors: etc/contrller master and scheduler are red and throw unhealthy errors.

Checking the pods, you'll find everything running expect the crashing cattle-node-agent:

vagrant@ubuntu-xenial:~$ kubectl get pods --all-namespaces
NAMESPACE       NAME                                    READY   STATUS             RESTARTS   AGE
cattle-system   cattle-cluster-agent-5f59578445-szlpw   1/1     Running            0          12m
cattle-system   cattle-node-agent-w5rj4                 0/1     CrashLoopBackOff   7          12m
cattle-system   rancher-6756fb68d4-bq7q6                1/1     Running            1          14m
cattle-system   rancher-6756fb68d4-gxdpz                1/1     Running            1          14m
cattle-system   rancher-6756fb68d4-s8pj4                1/1     Running            0          14m
kube-system     cert-manager-6464494858-qffzh           1/1     Running            0          15m
kube-system     coredns-7748f7f6df-c595c                1/1     Running            0          15m
kube-system     helm-install-traefik-m7gmj              0/1     Completed          0          15m
kube-system     svclb-traefik-7fd99b58f5-w9r87          2/2     Running            0          15m
kube-system     tiller-deploy-6bbdcdc7ff-92pwv          1/1     Running            0          15m
kube-system     traefik-dcd66ffd7-p27lx                 1/1     Running            0          15m

Error log:

vagrant@ubuntu-xenial:~$ kubectl logs -n cattle-system cattle-node-agent-w5rj4
ERROR: Please bind mount in the docker socket to /var/run/docker.sock
ERROR: example: docker run -v /var/run/docker.sock:/var/run/docker.sock ...

If we look into kubectl describe, it seems that volumes are mounted correctly:

   Mounts:
      /cattle-credentials from cattle-credentials (ro)
      /etc/kubernetes from k8s-ssl (rw)
      /run from run (rw)
      /var/run from var-run (rw)
      /var/run/secrets/kubernetes.io/serviceaccount from cattle-token-cpxw5 (ro)

To Reproduce
Install k3s using the official installer script, then install helm and rancher according to their docs. Wait a few minutes and check the pods as described above.

Expected behavior
Get cattle running like it does on the default k8s rancher cluster.

Additional context
The vagrant box image ubuntu/xenial64 (16.04.5 LTS) was used. I think this issue is related to containerd, which is used in k3s, where cattle expect the docker socket (which is not there). Additionally, I applied this dns workaround since I'm using a internal domain that couldn't be resolved from the pods otherwise.

[superseb]

This is already being covered in #69

[ibuildthecloud]

Closing in favor of #69 (This is actually already fixed on Rancher master rancher/rancher:master)