soft "start" without args - will overwrite the container-runtime to docker.

[medyagh]

in this PR I added an integeration test for Soft Start. in TestStartStop.
and it was revealed that our soft start is broken for not-docker runtimes

verified locally

verified locally this happens only when on the Second Start you do NOT specify the run time.

So:

minikube start --driver=hyperkit --container-runtime=conrtainerd
minikube start --driver=hyperkit --container-runtime=conrtainerd

(this one is fine)
but if you do

minikube start 

and in the config, you will see, the run time has been changed to docker

 $ cat ~/.minikube/profiles/minikube/config.json | grep "ContainerRuntime"
                "ContainerRuntime": "docker",

it is not fine !

---

Test Logs:

---

it appears that on a restart it generates a WRONG kubeadm config.

-  criSocket: /run/containerd/containerd.sock
+  criSocket: /var/run/dockershim.sock

https://storage.googleapis.com/minikube-builds/logs/7435/0f498bc/KVM_Linux.html#fail_TestStartStop%2fgroup%2fcontainerd

ing, string) (net.Conn, error))(nil)}
	E0406 03:53:37.937401   13793 kubeadm.go:331] Overriding stale ClientConfig host https://192.168.39.96:8444 with https://192.168.39.96:8443
	I0406 03:53:38.063752   13793 ssh_runner.go:101] Run: sudo diff -u /var/tmp/minikube/kubeadm.yaml /var/tmp/minikube/kubeadm.yaml.new
	I0406 03:53:38.118080   13793 kubeadm.go:404] needs reset: configs differ:
	-- stdout --
	--- /var/tmp/minikube/kubeadm.yaml
	+++ /var/tmp/minikube/kubeadm.yaml.new
	@@ -2,7 +2,7 @@
	 kind: InitConfiguration
	 localAPIEndpoint:
	   advertiseAddress: 192.168.39.96
	-  bindPort: 8444
	+  bindPort: 8443
	 bootstrapTokens:
	   - groups:
	       - system:bootstrappers:kubeadm:default-node-token
	@@ -11,7 +11,7 @@
	       - signing
	       - authentication
	 nodeRegistration:
	-  criSocket: /run/containerd/containerd.sock
	+  criSocket: /var/run/dockershim.sock
	   name: "containerd-20200406T034015.056085983-5464"
	   kubeletExtraArgs:
	     node-ip: 192.168.39.96
	@@ -25,7 +25,7 @@
	     enable-admission-plugins: "NamespaceLifecycle,LimitRanger,ServiceAccount,DefaultStorageClass,DefaultTolerationSeconds,NodeRestriction,MutatingAdmissionWebhook,ValidatingAdmissionWebhook,ResourceQuota"
	 certificatesDir: /var/lib/minikube/certs
	 clusterName: mk
	-controlPlaneEndpoint: 192.168.39.96:8444
	+controlPlaneEndpoint: 192.168.39.96:8443
	 dns:
	   type: CoreDNS
	 etcd:
	
	-- /stdout --
	I0406 03:53:38.118207   13793 ssh_run

also in
https://storage.googleapis.com/minikube-builds/logs/7435/0f498bc/KVM_Linux.html#fail_TestStartStop%2fgroup%2fcrio

ing, string) (net.Conn, error))(nil)}
	I0406 03:54:26.658488   14472 ssh_runner.go:101] Run: sudo diff -u /var/tmp/minikube/kubeadm.yaml /var/tmp/minikube/kubeadm.yaml.new
	I0406 03:54:26.715340   14472 kubeadm.go:404] needs reset: configs differ:
	-- stdout --
	--- /var/tmp/minikube/kubeadm.yaml
	+++ /var/tmp/minikube/kubeadm.yaml.new
	@@ -11,7 +11,7 @@
	       - signing
	       - authentication
	 nodeRegistration:
	-  criSocket: /var/run/crio/crio.sock
	+  criSocket: /var/run/dockershim.sock
	   name: "crio-20200406T034015.055332473-5464"
	   kubeletExtraArgs:
	     node-ip: 192.168.39.212
	
	-- /stdout --
	I0406 03:54:26.715449   14472 ssh_runner.go:101] Run: sudo /bin/bash -c "grep https://192.168.39.212:8443 /etc/kubernetes/admin.conf || sudo rm -f /et

[afbjorklund]

Is this the same driver registry bug that makes cri-o look for docker binary ? #5549