So, I am writing a custom auto rescheduler which should get node metrics (lets take memory usage in this case) and then if it is high reschedule a pod in another node with lesser memory usage. I know that cAdvisor picks up the metrics and gives it to heapster which are available at a rest endpoint. I don't have heapster deployed on my server and I use prometheus for scraping metrics. Is there any Kubernetes API which can be used to directly get node metrics, because I couldn't find any.