Infrastructure

What your cluster looks like right now: pods, nodes, and why any of them are unhappy.

Current health

Four counts across everything the platform can see.

Count	Means
Healthy	Running, ready, no recent restarts
Attention	Not ready, restarting, or under pressure
Stale	The last observation is too old to trust
Errors	The data source returned an error instead of a reading

Stale and Errors are about the platform, not your cluster. They mean the picture is unreliable, which is a different problem from the cluster being unhealthy, and is why they are counted separately.

Filters

Search matches the resource name. Status, kind, and namespace narrow the list. Whatever you filter to, issues are shown first, so the list never buries a crash loop under forty healthy pods.

Reading a row

Column	What it tells you
Health	The lane the resource is in
Resource	The name, and whether it is a pod or a node
Namespace	Its namespace, or cluster-scoped for a node
Signals	Why it is in that lane: the phase, the restart count, the container state, and any previous termination reason with its age
Observed	When this reading was taken, in your local time
Investigation	Opens an incident from this resource

The Signals column is the one that answers "why". A row does not just say "Attention"; it says CrashLoopBackOff, and it says the previous container was OOMKilled nine minutes ago.

Investigate

Investigate opens an incident from what the platform is looking at, for something monitoring has not alerted on. Because it was declared here rather than raised by an alert, it has no Slack thread and lives on the dashboard.

Where this comes from

flowchart LR
    Cluster["Your cluster"] --> Poll["The platform polls,<br/>read-only"]
    Poll --> Cache["Latest reading, held briefly"]
    Cache --> Page["This page"]
    Cache --> Triage["Evidence for an investigation"]

Only the most recent reading is kept, and only for a short time. This page is a live view, not a history, which is why a source that stops responding shows as Stale rather than quietly showing you yesterday.

To connect a cluster, see Kubernetes.