Second startup of all-in-one takes 100 seconds to finish starting #899
Description
On first startup, a node is auto created
On second startup, the node exists, and messages about the node already existing are printed 10 times, 10 seconds apart, until it gives up and continues.
start.go, 550: kmaster.RunMinionController()
kubernetes/master.go, 143: minionController.Run(10*time.Second, 10)
nodecontroller.go, 122: _, err := s.kubeClient.Nodes().Create(&node)
Looks like nodecontroller.go tolerated an "already exists" error until the latest upstream bump