Skip to main content
OpenCost is a CNCF project for Kubernetes cost monitoring. Kubecost builds on that model and can expose cost data through Prometheus metrics or a Kubecost 3 Aggregator API. Radar supports both paths. It can read OpenCost-compatible metrics through the already-connected Prometheus or current allocation and asset data from a Kubecost 3 Aggregator. No CRDs are required. Both integrations are passive and read-only.

​
What Radar shows

  • Current cluster rate and monthly projection.
  • CPU, memory, storage, and network cost when present.
  • Namespace and node breakdowns.
  • Focused application and workload CPU and memory allocation cost.
  • Historical 6h, 24h, and 7d trends with hover details. Kubecost supplies the retained cluster and namespace trend; workload and application history require the Prometheus path.
  • Separate CPU and memory allocation-use evidence when usage coverage is complete.
See Cost insights for the user-facing views and current scope.

​
Source selection and discovery

Auto mode prefers working OpenCost-compatible metrics from Prometheus, then tries a Kubecost 3 Aggregator in the connected cluster. You can pin either source under Settings → Cost, in config.json, or through Helm. For a local Aggregator, Radar discovers its active StatefulSet and matching Service. It tries named port tcp-api (9004) first. If that endpoint rejects unauthenticated access and no API key is configured, Radar can retry the same Service’s exact tcp-api-rbac port (9008) for SAML/OIDC-enabled Kubecost. An explicitly configured key is never bypassed. A Service exposing only port 9008 is not auto-discovered; configure that endpoint explicitly. Federated clusters that run only the Kubecost FinOps Agent require the central Aggregator URL and exact cluster ID. The API key is optional when the endpoint intentionally permits unauthenticated allocation reads. See Cost insights for the complete setup journeys, Settings behavior, and Helm values.

​
Rightsizing is separate

Rightsizing uses container demand metrics to provide workload-level request guidance. Cost data provides context, but neither OpenCost nor Kubecost produces Radar’s request suggestions.