// the find
prometheus-operator/kube-prometheus
Use Prometheus to monitor Kubernetes and applications running on Kubernetes
kube-prometheus is an opinionated, batteries-included Prometheus monitoring stack for Kubernetes, bundling Prometheus Operator, Alertmanager, Grafana, kube-state-metrics, node-exporter, and the Prometheus adapter into one deployable unit. It targets platform engineers who want cluster-wide observability without assembling the pieces themselves. The whole thing is written in Jsonnet, which is either a feature or a problem depending on whether you've used Jsonnet before.
The Jsonnet-as-library model is genuinely well thought out — you extend it by importing and patching, not by forking, so upstream updates don't require manual merge conflicts. The platform-specific addons (EKS, GKE, AKS, kubeadm) handle the annoying differences in how managed clusters expose control plane metrics, saving hours of debugging. Pre-baked alerting rules from kubernetes-mixin are solid starting points and cover the alerts most teams wish they'd had after their first outage. The Kubernetes compatibility matrix is clearly documented and CI actually tests against real clusters via kind, so the 'works on my version' problem is largely solved.
Jsonnet is a significant learning curve tax for anyone who just wants YAML — the toolchain (jsonnet, jb, build.sh) adds real friction before you get to the monitoring value. The repo's own README has a warning that 'everything is experimental and may change significantly at any time', which is not a great signal for a stack you're running in production. Customizing alert thresholds or Grafana dashboards requires understanding the mixin layering system, which is non-obvious and poorly documented outside of a handful of examples. The pre-generated manifests checked into the repo are fine for a quick try but they create a false sense that you can just `kubectl apply` and be done — production use really does require the Jsonnet path, and the gap between 'quickstart' and 'customized for your cluster' is steep.