Kepler: Energy Observability for Kubernetes
Shriira Press
Every watt your cluster burns can be measured, attributed, and acted upon — Kepler is how.
Welcome to Kepler: Energy Observability for Kubernetes.
Kepler, the Kubernetes-based Efficient Power Level Exporter, is a CNCF sandbox project that answers a question most monitoring stacks cannot: how much energy is each workload actually consuming? It reads real hardware energy counters — Intel and AMD RAPL zones exposed through the Linux powercap framework, GPU sensors, and platform power from a server's baseboard management controller — and attributes that energy down to individual processes, containers, virtual machines, pods, and nodes. The result is a stream of Prometheus metrics that turns vague sustainability goals into concrete numbers you can chart, alert on, and bill against. This book opens with the problem of invisible energy and the regulatory and cost pressures driving it, then builds up Kepler's architecture layer by layer: how it senses hardware power, how it tracks workloads through the proc filesystem and the Kubernetes API, and how its ratio model fairly splits idle and active power across everything running. From there it moves to the metrics catalog, deployment as a DaemonSet, Grafana dashboards and carbon estimation, and finally the practical art of running Kepler well and reading its numbers honestly. By the end you will understand not just what Kepler reports, but why those numbers mean what they mean.
This title is part of the ShriIra library and is free to read in full, right here — our small contribution to making world-class knowledge easy to reach.
A note on reading it: open the Contents menu at the top of the reader to jump between chapters, use the Aa menu to set a comfortable text size, theme (light, sepia, or night), and single- or two-page layout. Your place is saved automatically, so you can always pick up where you left off.
We hope it serves you well.
— Shriira Press