Skip to content

Metrics & health

Kubus reads CPU and memory from metrics-server and keeps a short rolling history, so you get little trend charts rather than a single instantaneous number.

CPU and memory history charts for a pod CPU and memory history charts for a pod
Live CPU/memory history right in the details drawer.

Where you'll see metrics

  • Metrics page — the sidebar entry below Topology: cluster-wide CPU/memory trends, per-node lines, top pods, and a per-namespace breakdown for every selected cluster. If a cluster has no metrics-server yet, the page offers a one-click install instead.
  • Details drawer → Metrics tab — for Pods and Nodes, live CPU/memory charts with the current value called out.
  • Resource lists — CPU and memory columns on the Pods list.
  • Overview dashboard — a per-node usage table. More →
The cluster-wide Metrics page The cluster-wide Metrics page
The Metrics page: cluster trends, per-node lines and top pods.

History accumulates while Kubus is open, so the first samples appear within ~20 seconds of opening a chart and fill in from there.

Installing metrics-server from Kubus

If a cluster has no metrics-server, Kubus can install it for you — no kubectl or Helm required:

  • On the Overview page, the Node usage card shows an Install metrics-server button whenever usage data is unavailable.
  • The install dialog has one option: Skip kubelet TLS verification (--kubelet-insecure-tls). Enable it on local/dev clusters — kind, minikube, docker-desktop — whose kubelets serve self-signed certificates.

Kubus applies the official pinned components.yaml (Deployment, Service, RBAC and the metrics.k8s.io APIService in kube-system) via server-side apply, and labels the resources as managed by Kubus. Graphs appear within a minute of the pod becoming ready. Re-running the install is safe — it re-applies the same manifest, which also repairs a broken install.

To remove it again, use the Uninstall button in the Metrics page header. If the metrics-server wasn't installed by Kubus (k3s bundles one; cloud distributions often manage their own), Kubus warns you first — your distribution may recreate it or expect removal through its own tooling.

Prefer the CLI? The equivalent manual install:

kubectl apply -f https://github.com/kubernetes-sigs/metrics-server/releases/latest/download/components.yaml

No metrics-server? No problem.

Without it, Kubus just hides the charts and shows "Metrics unavailable" — every other feature works normally.

Controlling refresh

Metrics, events and the overview poll on a timer. You can speed that up or dial it back (or turn polling off) under Settings → Data & refresh. Watched lists stay live over their WebSocket regardless of this setting.

See also