Dashboard
A bird's-eye view of the cluster — nodes, pods, workloads, capacity and problems — where every number opens what is behind it.
Loupe opens on the dashboard once a cluster connects. It answers "how is this cluster doing?" before you have chosen where to look.
What it shows
- Nodes — ready and cordoned
- Pods by phase, with crash loops called out
- Workloads healthy by kind — Deployments, StatefulSets, DaemonSets
- CPU and memory — live usage when metrics-server is installed; otherwise requests against allocatable, and the widget says that is what it is showing
- Pod slots — how close the cluster is to each node's max-pods
- Busiest nodes, the containers restarting most, namespaces by pod count and volume-claim status
- Problems — the Problems list as a widget
Every number opens what is behind it — the listing or object it counts. ⌘-click opens it in a new tab.
Cost to the cluster
The counts ride on the same watches that keep Problems
current, so the dashboard is live and makes no extra LIST requests.
Live usage comes from metrics.k8s.io when it is served.
If RBAC refuses one of the sources, its widget says so rather than showing zero — a dashboard that reads healthy because it could not look is worse than none.