CNCF Atlas

English

日本語

English

日本語

Appearance

Getting Started ​

Verified against the documented commit 4d117aa. Commands assume a Kubernetes 1.20+ cluster with Prometheus, per README.md:33-46.

Prerequisites ​

  • A Kubernetes cluster, version 1.20 or newer.
  • Prometheus running in the cluster (OpenCost queries it for usage metrics).
  • Helm, the only supported install method. The standalone manifests have been removed (README.md:33).

Install ​

bash
helm repo add opencost https://opencost.github.io/opencost-helm-chart
helm repo update
helm install opencost opencost/opencost

A first working setup ​

The shortest path that actually computes cost. The Helm install above is the recommended route. To run the engine locally without deploying it, port-forward to Prometheus and start the binary against it.

  1. Forward the in-cluster Prometheus service to your workstation.

    bash
    kubectl port-forward svc/prometheus-server 9080:80

  2. Point OpenCost at that endpoint and run the cost model from source.

    bash
    PROMETHEUS_SERVER_ENDPOINT="http://127.0.0.1:9080" go run ./cmd/costmodel/main.go

The API listens on port 9003 by default.

Verify it works ​

Query the allocation API for a window. The window parameter is required (pkg/costmodel/aggregation.go:337).

bash
curl "http://127.0.0.1:9003/allocation?window=1d&aggregate=namespace"

A healthy setup returns a JSON allocation set keyed by namespace. Other endpoints are /allocation/summary, /assets, /cloudCost, and /metrics.

Where to go next ​

  • For sharded or HA Prometheus, set PROMETHEUS_SERVER_ENDPOINT to a global query endpoint such as Thanos Query, Cortex, or Mimir; pointing at a single Prometheus pod gives incomplete results (README.md:46).
  • The OpenCost documentation covers Helm configuration, the Prometheus integration, and the UI for production concerns not repeated here.