vagrant based cloud init / userdata will fail on subsequent reboots due to missing values

[aaronlevy]

When using vagrant, we populate an env file which defaults to a /run/... path. On subsequent reboots - that file will no longer exist and the user data scripts fail (here: https://github.com/coreos/coreos-kubernetes/blob/ea3bd80f9cc380df67b7c20033b44f1f7c79d5a0/multi-node/generic/controller-install.sh#L45)

This doesn't cause a problem because the node has already been configured - but we should make sure we're not relying on this failure to hide other places where we stomp on configuration after reboot.