Skip to main content

Documentation Index

Fetch the complete documentation index at: https://radarhq.io/docs/llms.txt

Use this file to discover all available pages before exploring further.

If OpenCost is installed, Radar surfaces a Cost view that breaks down spend by namespace, workload, and node, with trend lines and an efficiency score. No separate dashboard to set up - Radar reads OpenCost’s opencost_container_*_cost_total series from your Prometheus.

​
What you see

  • Top-line - total cost in the selected window, plus the change vs. the previous equivalent window.
  • By namespace - sortable, with breakdown into CPU, memory, GPU, storage, and network.
  • By workload - same axes, scoped to Deployments / StatefulSets / DaemonSets / Jobs.
  • By node - per-node cost, utilization, and idle %.
  • Efficiency - requested vs. used, surfaced both as a workload-level score and a cluster-level rollup.
  • Trends - cost over time (last 24h / 7d / 30d / custom).

​
How it auto-detects

Radar reads OpenCost via Prometheus - if the Prometheus auto-discovery (or manual traffic.prometheusUrl / --prometheus-url override) returns opencost_container_cpu_cost_total and opencost_container_memory_cost_total series, the Cost view appears in the sidebar. If those series aren’t queryable, the Cost view is hidden.

​
Common gaps

If OpenCost is misconfigured (no cloud provider integration, missing on-prem rate card), you’ll see CPU and memory dollars but not GPU, network, or persistent volume costs. Fix OpenCost and Radar picks up the new series automatically.
  • GPU pricing - needs a GPU pricing config in OpenCost.
  • Cloud egress - needs cloud-provider integration credentials.
  • Persistent volumes - needs storage-class to price mapping.
  • Spot vs. on-demand - needs node labels OpenCost can read.

​
RBAC

Radar’s ServiceAccount queries Prometheus via the in-cluster Service. Default rbac.create: true is sufficient - no extra permissions needed.

​
Limitations

  • No write actions. Radar reads OpenCost; it doesn’t change instance types or evict workloads. We’re not a cost remediation tool.
  • No multi-cluster aggregation in OSS. Costs are per-cluster. Radar Cloud rolls them up across the fleet.

​
See also