The problem

Node-level power tells you a cluster is expensive, not which service to fix. In shared clusters, one team's workload hides inside everyone's bill.

The fix you can use today

• Deploy a per-pod energy exporter and send it to your existing dashboards.
• Rank namespaces and services by energy per request.
• For a clean benchmark, pin the workload to a dedicated node.

The Kepler project measures the energy of each pod in a Kubernetes cluster. GreenCode's research review looks at tools like this.

Tools that help

• Kepler: exports energy estimates per pod and container on Kubernetes, using eBPF and hardware counters
• Scaphandre: measures the power drawn by each process on a Linux server and exports it to Prometheus and other monitoring stacks

How GreenCode helps

GreenCode helps tackle this by tracing energy and CO2e to the services that cause it, including services running hot and services barely running at all. It mines a running estate and rebuilds it as infrastructure as code. More on this.

Tags

#kubernetes#measurement

Sources

• GreenCode State of the Art Review, D2.1 (2025), PDF p. 26: Kepler and Cloud Carbon Footprint attribute energy to pods, containers or cloud accounts
• GreenCode State of the Art Review, D2.1 (2025), PDF p. 28: Kepler: pod and container-level estimates, cluster-wide metrics

All GreenCode Tech Tips