Technology · Ebook

Kepler: Energy Observability for Kubernetes

by Shriira Press

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 — RAPL zones via the Linux powercap framework, GPU sensors, and platform power from a server's BMC — and attributes that energy down to processes, containers, VMs, pods, and nodes as Prometheus metrics. This book builds Kepler up layer by layer: the problem of invisible energy, the hardware power sources, how it tracks workloads through proc and the Kubernetes API, its idle-versus-active ratio attribution model, the metrics it exports, the v0.10 service-oriented architecture, deployment as a DaemonSet with Grafana and carbon estimation, and the practice of reading its numbers honestly.

Contents

1. 1Preface
2. 2Chapter 1 — The Energy You Cannot See
3. 3Chapter 2 — Where Power Comes From
4. 4Chapter 3 — Watching the Workloads
5. 5Chapter 4 — Attributing Energy Fairly
6. 6Chapter 5 — The Metrics Kepler Exports
7. 7Chapter 6 — Architecture of the Exporter
8. 8Chapter 7 — Deploying and Visualizing Kepler
9. 9Chapter 8 — Kepler in Practice