// the find
kubernetes-sigs/kubernetes-mixin
A set of Grafana dashboards and Prometheus alerts for Kubernetes.
The upstream Grafana dashboards and Prometheus alerting/recording rules for monitoring Kubernetes, written in Jsonnet and maintained under kubernetes-sigs. It's for anyone running Prometheus + Grafana on Kubernetes who doesn't want to hand-write RED-method dashboards and alerts from scratch.
The Kubernetes-version compatibility matrix is unusually rigorous — it names the exact metrics that changed (apiserver_request_sli_duration_seconds in 1.26, scheduler_pod_scheduling_sli_duration_seconds in 1.29) and which PRs introduced them, instead of a vague 'supports 1.28+'. CI actually runs promtool-style rule tests against the generated YAML (tests/*.yaml), not just a jsonnet syntax check. The override pattern in config.libsonnet lets you rename selectors, dashboard prefixes, and tags by composing a local libsonnet file rather than forking the repo. Multi-cluster support via Thanos/Cortex external labels is a first-class config flag, not bolted on.
Jsonnet is a hard requirement just to get usable output — you need jb, the jsonnet compiler, and to understand mixin composition before you have a single YAML file to deploy. The prometheus-operator integration path, arguably the most common way people run this stack today, is a bare 'TODO' in the README. There's a recording rule flagged deprecated in-repo (container_cpu_usage_seconds_total:sum_irate) that keeps shipping until v2.0.0 with no automated migration warning if you're still consuming it. The maintainers are upfront that the compatibility table doesn't guarantee anything works with your specific exporter versions — automated metric-compatibility checking is still an open issue (#1249), so silently-broken panels after a k8s upgrade are still your problem to catch.