What it does
kube-state-metrics is a Kubernetes add-on that reads API object state and exposes cluster-level metrics. Its measurements describe objects such as pods and their status or phase, rather than operating as a complete monitoring service. Another system must scrape the endpoint and provide storage, querying, dashboards, or alerting.
Turn object state into an operational signal
kube-state-metrics watches Kubernetes objects and exposes state-derived samples through a metrics endpoint. That can add context a resource-usage chart does not provide, such as whether a pod or other object is in the expected state. A practical first use is to deploy the exporter with the cluster's monitoring stack and have that stack scrape its endpoint; dashboards and alerts are then built in the downstream system. No deployment or cluster scrape was performed for this research.
Keep its scope clear
The exporter reflects the current state of Kubernetes objects. Its project documentation notes that an object's metrics stop appearing after that object is deleted, so it is not a durable event history. kube-state-metrics also does not provide the time-series database, query interface, or notification workflow itself. The distinction matters when choosing a first step: use it to expose object-state signals, and rely on a separate monitoring backend to retain and interpret them.
The project's Apache 2.0 license is available in its official repository. The repository publishes a Kubernetes compatibility matrix; check it against the cluster version and current release before deployment. This record describes documented behavior only, not a tested install.
Strengths
- Exposes metrics derived from Kubernetes API object state, including object status and phase information.
- Its metrics endpoint can feed a separate scraper and monitoring backend for dashboards and alerts.
- The project is licensed under Apache 2.0.
Watch out for
- Metrics describe current object state, and the project's README notes that metrics for deleted objects no longer appear at the endpoint.
- It does not provide durable metric storage, queries, dashboards, or alerts; those require another monitoring system.
Best for
- Kubernetes platform teams adding object health and status measurements to an existing monitoring stack.
- Operators building queries or alerts around workload state rather than only resource consumption.
Sources
- LICENSE (Oct 26, 2026)
- kubernetes/kube-state-metrics README (Oct 26, 2026)