- when performing transactional-update from CLI another
  configuration is used and afterwards the file
  /var/run/reboot-needed is created, instead of
  /var/run/reboot-required and Kured does not work
  as expected
- the upgrade has to be triggered using the systemd
  service instead

- it's important to switch off auto upgrading completely
  in the kube.tf and not just by removing the node labels
- otherwise the labels are created again on each agent
  restart or node reboot
- the previous wording implied that the labels are only
  an alternative to switching off, but they are actually
  an alternative to switching on.

- path has to point to the host's /var/run mountpoint
  in the container