The cluster page¶

Open a cluster from Clusters or the overview for everything ARGUS knows about it, one tab per question.

A cluster's overview tab: rings for pods, workloads and nodes, resource use against
allocatable, objects by kind, what needs attention and recent changes

The header carries the cluster's state, last heartbeat, Kubernetes and agent versions. Tabs with something to look at carry a count.

The header¶

The header shows the cluster's state, when the agent last reported and when it last sent its facts, the Kubernetes version, the agent version and whether it is compatible with this hub, and when the cluster was enrolled.

Tabs¶

Tab	Answers
Overview	Is this cluster healthy, and what needs attention
Timeline	What happened here, and when
Incidents	Which of this cluster's incidents need someone
Map	How workloads connect: entry, service, workload, pods
GitOps	What Argo CD and Flux report about their applications
Helm	Which releases are deployed and how healthy their objects are
Certificates	What expires next
Upgrade	What would break on the next Kubernetes version
Posture	Which settings carry security, reliability or efficiency risk
Traffic	Which workloads talk to which, from a traffic source in Prometheus
Network policy	What may talk to what, and where policy is missing or too open
Cost	Spend per namespace and its trend, from OpenCost or Kubecost, and rightsizing per container
Agent	What the agent watches, how detection is running, and what it may change

The selected tab is part of the address, so a link opens the same view.

Overview¶

Rings for pods, workloads and nodes, each split into words — running, completed, failed, ready, not ready — and Resource use: CPU and memory used and requested against what is allocatable. Objects by kind counts what is in the cluster. Needs attention lists open incidents and problems by severity, and Changes charts recent changes per five minutes, with the busiest objects listed. On a cluster running Karpenter, Karpenter capacity shows each NodePool's limits against what its nodes use, and any NodeClaims not yet ready.

Timeline¶

The timeline tab: incidents opened in the window with what changed before each, and a
strip per object of changes, findings and incidents

Each incident shows what changed before it opened. Below, one row per object marks its changes, its findings and its incidents on one time axis.

Choose a window — one hour to seven days — and narrow by namespace or type. Objects with an incident come first.

Incidents¶

The cluster's incidents tab: open incidents by severity and by lifecycle step, then the
list with look-alike incidents grouped

This cluster's incidents, with the open set split by severity and by where each incident is in the lifecycle. Look-alike incidents — the same fault on many pods — are grouped into one row. Open incident worklist opens the full incident list narrowed to this cluster.

Agent¶

The agent tab: watch coverage, the object cache, detector status, remediation authority
and compatibility

What the agent can see, how its detectors are running, and what it may change.

Card	Shows
Watch coverage	How many resource kinds the agent watches, and which it cannot read
Object cache	Which kinds are synced, forbidden or not installed. Kinds that are not installed or not readable are looked up again every few minutes, so installing a controller or granting a read later needs no agent restart
Detection	Each detector's status and last run, the sustain period, resend cooldown, detection latency and finding queue
Remediation authority	Each write action and whether the agent holds it — discovered by asking the cluster, not taken from configuration
Metrics source	Whether the agent can query Prometheus, in-cluster or an external endpoint. When none is configured, the Prometheus, Thanos or VictoriaMetrics Service it found, with the chart values to set. An external endpoint is marked Not scoped to this cluster when no `extraLabel` is set, and Label not enforced when one is set but the endpoint returned series outside it, as plain Prometheus and Thanos do
Compatibility	Whether this agent version is supported by the hub, and the image it runs. When the hub ships a newer agent image and the chart allows it, Upgrade agent moves both agent Deployments onto it; otherwise the card shows the `helm upgrade` command

Remediation authority is answered by the remediation identity itself, so it describes what an approved plan can actually change. It has three states: not reported (the agent predates the check), not answered (it asked and got no answer, which is not a denial) and a real per-action answer. With no remediation agent connected, the card says so. If an approval fails to execute, check here first.