Atlas
A live map of how everything connects — ownership chains and the references between them, drawn from the cache and navigable by click.
The Atlas draws the cluster as it actually connects: ownership chains (Deployment → ReplicaSet → Pod) and the references that tie them together — Services selecting workloads, Ingresses routing to Services, volumes and config feeding pods. It's built from the daemon's live cache, so it redraws as the cluster moves, and every node is a door: click through to the inspector for that object.
It's the page for the questions a list can't answer — what fronts this Service? what shares this volume? what will this delete take down with it? — and the same map powers the Related tab on every object.
Keeping big clusters drawable
A thousand-node hairball helps nobody, so what Atlas draws is yours to shape in Settings → Atlas display:
- Hide kinds you don't care to see.
- ReplicaSets are folded out of the chain by default — Deployment connects straight to Pods — and can be shown when the intermediate matters.
- A node budget caps how much one drawing attempts, so a huge namespace degrades to something legible instead of something slow.
These are per-cluster settings: the map of your busy production context and the map of a three-pod kind cluster can be tuned separately.