minikube start hangs when addon-manager is disabled

[andahme]

When addon-manager is disabled, minikube start hangs for about 10m before timing out. After the timeout, the cluster is in a working state as desired.

• The exact command-lines used so that we can replicate the issue

minikube start

• The full output of the command that failed

o   minikube v0.35.0 on darwin (amd64)
>   Creating virtualbox VM (CPUs=2, Memory=2048MB, Disk=20000MB) ...
-   "minikube" IP address is 192.168.99.100
-   Configuring Docker as the container runtime ...
-   Preparing Kubernetes environment ...
-   Pulling images required by Kubernetes v1.13.4 ...
-   Launching Kubernetes v1.13.4 using kubeadm ... 
:   Waiting for pods: apiserver proxy etcd scheduler controller addon-manager

*   Sorry that minikube crashed. If this was unexpected, we would love to hear from you:
-   https://github.com/kubernetes/minikube/issues/new

• Which operating system version was used
  macOS 10.4.3

• Which Minikube version was used
  v0.35.0

[tstromberg]

I agree that we can do better here. Any particular reason for disabling the addon manager?

[andahme]

In particular, I have been modifying ingress addon resources (see comments in 2922) which have a Reconcile mode that makes this tricky.

I did recently discover the tunnel feature, which may change how I do ingress in Minikube. I would make a broader case.

Minikube is my primary k8s experimenting/learning/scratch environment and a test bed for things I plan to take to other environments. The ability to standup a minimal "vanilla" cluster is pretty awesome.

[andahme]

I think the only real issue here is a user experience one. I downgraded Minikube and reset config (deleted ~/.minikube) trying to figure out why (it appeared) I could not start a cluster anymore. In the alternate case (did not hang), I would have realized I needed to enable addon-manager again to lay down the base ingress addon resources (and storage-provisioner/etc).

Since addon-manager can be disabled and it doesn't appear to be a requirement for cluster operation, I would think it would be ok/good to remove it from the startup wait list.