g00x/idlekube

IdleKube is a lightweight command line tool for looking at a Kubernetes cluster and pointing out where compute resources are being wasted. It checks for workloads that are barely being used, containers whose CPU or memory requests are much higher than what they actually consume, and deployments that are missing ownership labels. It pulls this information from the Kubernetes API and from live usage data provided by metrics-server, then puts together a quick snapshot of where waste is likely happening. The tool is explicit about what it is not. It is not a billing system, it does not automatically resize your workloads, and it is not meant to replace dedicated cost tools like Prometheus, OpenCost, or Kubecost. Instead, it helps you figure out where to look first when doing a cleanup or an audit, ranking namespaces and workloads by potential monthly and annual savings and flagging the top candidates to review. To use it, you clone the repository, set up a Python virtual environment, install the dependencies, and run a scan command against your cluster. Results can be viewed on screen or exported as JSON, CSV, or HTML reports, which makes it easy to share findings or feed them into other tools. You can scan the whole cluster or filter down to a single namespace, and adjust the assumed cost per CPU and memory unit to match your own cloud pricing. Because it works from a single snapshot of usage rather than historical data, its suggested review targets are described as low confidence, and the README recommends validating any change with at least a week or more of real usage history before adjusting resource requests or limits in production. It requires Python 3.10 or newer, a working kubectl setup, and metrics-server installed in the cluster. This suits developers and platform engineers who manage Kubernetes clusters and want a fast first pass at spotting overprovisioned or idle workloads.