Currently Dashboard only integrates metrics-server as a metrics source. It is using integration framework that allows to support and integrate more metric providers as well as additional applications such as Weave Scope or Grafana.
Metric integrations allow Dashboard to show cpu/memory usage graphs and sparklines of resources running inside the cluster. In order to make Dashboard resilient to metric provider crashes there was --metric-client-check-period flag introduced. By default every 30 seconds health of the metric provider will be checked and in case it crashes metrics will be disabled.
For the sparklines and graphs to be shown in Dashboard you need to have metrics-server running in your cluster. It now uses dashboard-metrics-scraper that is deployed by default with Kubernetes Dashboard. It uses the Metrics API to gather metrics.
The easiest way to check if metrics-server is installed and working properly is to run kubectl top pod or kubectl top node.
Copyright 2019 The Kubernetes Dashboard Authors