Graph and Live Views

Every view here updates live while it’s open, and stops watching when you leave it.

Graph (g)

For Kustomizations, HelmReleases, ResourceSets, FluxInstances, and ArtifactGenerators: the sources upstream and everything the resource manages downstream. Managed objects are grouped: Flux resources a ResourceSet produces appear as their own nodes, workloads gather in a workload group, and everything else in a resource group counted by kind.

• j / k move the focus between nodes; the view follows it.
• Enter opens the focused node; y / d show its YAML / describe output.
• Esc returns to the graph from anything you opened there.

Workloads

Enter on a workload group lists its Deployments, StatefulSets, and DaemonSets, with live readiness. Enter on one opens its workload detail:

• rollout status, containers and images, pods with restarts, and events
• Resources: each pod’s CPU and memory next to its request and limit, with a bar against the limit that turns amber at 75% and red at 90% (throttling / OOM risk). The source (kubelet, metrics-server, or requests only) is detected automatically — see Pod usage metrics.
• l streams a pod’s logs (a picker opens when there are several), r rolls out a restart, Ctrl+d deletes a pod so its controller replaces it. Restarts and deletes ask first.

Inventory

Enter on a resource group lists every object it holds with live kstatus health: Current, InProgress, Failed, Terminating, or NotFound / Forbidden when it can’t be read, with a message for anything short of Current. Enter / d describes an object, y shows its YAML (Secret values always redacted).

Trace (t)

What manages a resource, step by step: its controller owners, the Flux object that applies it (Kustomization, HelmRelease, ResourceSet, or FluxInstance), and that object’s source.

t traces the selected resource; :trace <kind>/<name> traces one by name in the current namespace (default when watching all namespaces):

• Flux kinds and aliases: :trace ks/apps, :trace hr/podinfo
• Exact Kubernetes kind names: :trace Deployment/web
• Plurals and short names (:trace deploy/web, :trace sts/db) need native resources on

HelmRelease values (v)

The values Helm actually receives: every spec.valuesFrom ConfigMap and Secret merged in order, then spec.values on top, exactly as helm-controller builds them.

• Each source shows whether it was applied, skipped (optional and absent), or failed (missing, forbidden, invalid YAML); one failure doesn’t hide the rest.
• targetPath follows Helm --set typing.
• Values from a Secret show as <redacted> until x reveals them.
• / searches, and R still reconciles the release.

Events (:events)

A live feed of Kubernetes Events in the current namespace (or the whole cluster after :ns all), newest first, Warnings highlighted. Flux emits an event for every reconciliation, so this shows error detail the list’s MESSAGE column truncates.

• / filters by type, reason, object, namespace, source, or message.
• Enter opens the event’s object; y / d work for any kind, and Flux keys (t, g, s, R, …) work when it’s a watched Flux resource.

Each resource’s describe view (d) also ends with its recent events.

Pulse (:pulse)

“Is my GitOps pipeline healthy?” at a glance: ready / failed / suspended totals per kind, the latest failures with their messages, Flux version and install status from the FluxReport, and controller pod health.

Controller logs (:logs)

Stream a Flux controller’s logs: :logs picks a pod, :logs <pod> streams one by name or prefix. Scrolling up pauses following and G resumes it, / searches, and c switches container (the default container comes from the kubectl.kubernetes.io/default-container annotation, as kubectl logs does).