Currently, the cluster setup script will try to get the docker info without checking if the command exists in the machine, and tries to continue to do the rest. Don't ask me why I know this 
EDIT: It also fails to verify if the docker daemon is not running, but checking that may not be worth it