Footprint
The Footprint pages give the supporting context for optimization: what each workload, namespace and node costs and emits. Recommendations and measured savings live elsewhere (Web UI). The figures come from the Impact Model; how energy and CO₂e are estimated is explained in Energy & Carbon.
Footprint data is not exported to Prometheus: use these pages, their exports and the API below (Prometheus & Grafana).
Overview (/footprint)
Section titled “Overview (/footprint)”KPIs and charts of the selected period and namespace: CO₂e, cost, energy and pods. Cluster-wide, the page shows three figures side by side for CO₂e, cost and energy, never added together:
| Figure | Meaning |
|---|---|
| Allocated | What the workloads (pod rows) are charged |
| Node idle share | Unreserved node capacity (the __idle__ rows) |
| Cluster total | Allocated + idle |
When the idle share is unknown for the period (for example rows older than the first idle row), the page says “not available” with the reason, never zero.

/footprint/nodes lists the node inventory (GET /api/v1/nodes) with CPU and memory capacity. A node whose GPU power is estimated from its TDP (gpu_tdp_estimate) or not counted (unmeasured), because no DCGM exporter sample reached GreenKube, shows a “GPU power estimated” or “GPU power not measured” badge, and a notice counts such nodes.
Report
Section titled “Report”
/footprint/report configures, previews and exports the cost, energy and CO₂e report as CSV or JSON. The old /report and /nodes routes redirect to /footprint/report and /footprint/nodes.
Every exported row has a row_kind column: allocated for workload rows (pods, or namespaces when grouped by namespace). By default the file holds only these rows, so a column sums to the report summary total. With include idle (include_idle=true) on a cluster-wide export, each selected range ends with:
- one
idlerow (pod_name=__idle__, the node idle share); - one
cluster_totalrow (pod_name=__cluster_total__, allocated + idle).
Do not sum every row of such a file: cluster_total already equals the allocated rows plus idle. Their values are empty (CSV) or null (JSON), never zero, when the idle share is unknown. Exports keep operational (co2e_grams) and embodied (embodied_co2e_grams) emissions in separate columns.
| Method | Path | Description |
|---|---|---|
GET | /api/v1/metrics/summary | Totals for a period (last, optional namespace); cluster-wide, adds idle_* and cluster_total_* |
GET | /api/v1/metrics/timeseries | Time series (granularity=hour|day|week|month) with the same idle and cluster-total fields |
GET | /api/v1/metrics/by-namespace, /metrics/top-pods | Per-namespace figures, highest-emitting pods |
GET | /api/v1/nodes | Node inventory, with each node’s gpu_source |
GET | /api/v1/report/summary | Report row count and totals; cluster-wide, adds the idle share and the cluster total |
GET | /api/v1/report/years | Calendar years with reportable data |
GET | /api/v1/report/export | Stream the report (format=csv|json, start/end or last, namespace, aggregate, granularity, group_by=namespace, include_idle) |
# Monthly report per namespace for a year, as CSVcurl -OJ -H "Authorization: Bearer $GREENKUBE_API_KEY" \ "http://localhost:8000/api/v1/report/export?start=2025-01-01&end=2025-12-31&aggregate=true&granularity=monthly&group_by=namespace"
# The last 30 days of one namespace, as JSONcurl -OJ -H "Authorization: Bearer $GREENKUBE_API_KEY" \ "http://localhost:8000/api/v1/report/export?format=json&last=30d&namespace=production"