First Prometheus Operator

K8s-native install.

Overview

The Prometheus Operator turns Kubernetes monitoring into declarative configuration. ServiceMonitor and PrometheusRule CRDs replace config-file scraping; teams own their own scrape configs and alert rules without touching the central Prometheus deployment.

• K8s-native install. Helm chart or kube-prometheus-stack bundle. Standard install pattern.
• ServiceMonitor CRD. Declarative scrape configuration as a Kubernetes object. Each service ships its own.
• PrometheusRule CRD. Alert rules as Kubernetes objects. Version-controlled alerts in the same Git repo as the service.
• Auto-discovery plus bundled Grafana. ServiceMonitors are discovered automatically; the stack includes Grafana with default dashboards.

The approach

Three habits keep declarative monitoring sustainable: install via the stack chart, distribute ServiceMonitors per service, and let each team own its own PrometheusRules.

• kube-prometheus-stack. Standard Helm install includes Prometheus, Alertmanager, Grafana, and node-exporter. Avoid hand-rolled installs.
• ServiceMonitor per service. Each service ships its own scrape config alongside the deployment manifest. Team autonomy without central friction.
• PrometheusRule per team. Each team owns its alerts. Cross-team alerts live in a shared rule set with clear ownership.
• Resource requests plus documented topology. Prometheus needs significant memory; the operator setup per cluster lives in the runbook.

Why this compounds

The first install is the foundation. Each new service plugs into the auto-discovery; team visibility compounds with every ServiceMonitor without operator intervention.

• Visibility scales. Automatic discovery produces broad coverage as new services ship. No central bottleneck.
• Alert hygiene. Version-controlled PrometheusRules support evolution. Bad alerts get reverted via Git.
• Reusable templates. Standard ServiceMonitor patterns scale across teams. New services copy from a working example.
• Year-one investment, year-two habit. The first install is heavy lift. By year two, every new service ships with monitoring from day one.