With kubernetes/node-problem-detector#29, node problem detector will not use hostNetwork any more.

However, when I start new node problem detector in the cluster, I found that the pod on master can't access apiserver via "10.0.0.1:443" - Get https://10.0.0.1:443/api/v1/namespaces/kube-system/pods/node-problem-detector-v0.2-lqewd: dial tcp 10.0.0.1:443: i/o timeout.
The pods on nodes work well.

Is this WAI?
/cc @kubernetes/sig-network