Kubernetes gives you a lightweight, read-only view of any clusters the agent finds in your fleet. It is not a full kubectl replacement — think of it as an at-a-glance sanity check next to everything else HostAtlas monitors.

What you see

At /kubernetes each cluster is a card:

• Host running the control plane, plus a healthy / degraded badge.
• Nodes — ready over total.
• Pods — running over total.
• CPU and Memory — cluster-wide averages.
• Node grid — one small square per node, coloured by ready state.
• Problem pods banner — appears if any pods are in a non-Running / non-Succeeded state, with the first couple of names.

Open a card to drill into a single cluster.

Cluster detail

For one cluster:

• Summary — control-plane host, kubectl version, when the data was last collected.
• Nodes table — name, status, roles, kubelet version.
• Pods by namespace — grouped list showing pod name, status, restarts, node.
• Deployments by namespace — desired vs. available replicas.

What you can do

Today, this page is read-only. There are no in-app actions to scale, restart, or reschedule.

How it works

• Detection is automatic. If a monitored host has kubectl available and can reach a cluster, the agent collects node, pod, and deployment metadata on its regular pass.
• Data is a snapshot, not a live stream. The last-collected timestamp is shown on each card.
• Only clusters reachable from a monitored host appear — HostAtlas does not talk directly to Kubernetes API servers.

Note

Deeper Kubernetes UX (log streaming, exec, describe, YAML editing, per-pod metrics) is on the roadmap. Until then, use the current view as a “did anything obvious break” check and keep kubectl for real work.

Related

• Docker Containers — non-K8s container inventory.
• Servers — the underlying hosts running your clusters.