To get a better understanding of how my Kubernetes cluster is actually doing and how hot the Rpi nodes are, I added some monitoring. Hats off to Carlos Eduardo for his hard work to enable us to install Prometheus and Grafana for ARM64 platform.

The instructions for installation are in Carlos’ repository. All that is required is to edit vars.jsonnet and then run the installation commands. Here are a couple of variables from vars.jsonnet you should keep an eye on:

name: 'armExporter',
      enabled: true,
  // Persistent volume configuration
  enablePersistence: {
    // Setting these to false, defaults to emptyDirs.
    prometheus: true,
    grafana: true,
    // If using a pre-created PV, fill in the names below. If blank, they will use the default StorageClass
    prometheusPV: '',
    grafanaPV: '',
    // If required to use a specific storageClass, keep the PV names above blank and fill the storageClass name below.
    storageClass: 'longhorn',
    // Define the PV sizes below
    prometheusSizePV: '20Gi',
    grafanaSizePV: '20Gi',
  },

I had some difficulties with finding the right size for PVs and their retention. After a bit of experimentation I was content with 20GB for both of them with 5 day retention.

When the installation is ready, log in to Grafana and search for dashboard named “Kubernetes cluster monitoring (via Prometheus)”. This is a data-rich dashboard from Carlos that shows you all the essential data at once.