Prometheus Operator
The orchestration brain. Manages Prometheus instances using native Kubernetes CRDs — ServiceMonitors, PodMonitors, and PrometheusRules for declarative configuration.
Core Engine Production-Ready Monitoring Stack
Kube Prometheus Stack
Deploy Prometheus, Grafana, Alertmanager, and essential exporters as a unified Helm chart. Enterprise-grade monitoring for your Kubernetes clusters — configured in minutes, not days.
0
GitHub Stars
0
Helm Downloads
0
Uptime SLA
Core Components
Six Pillars of Full-Stack Monitoring
Every component works in concert to deliver end-to-end observability — from infrastructure metrics to intelligent alerting and beautiful dashboards.
Prometheus Server
Industry-standard time-series database. Scrapes, stores, and evaluates metrics from your entire cluster with powerful PromQL query language support.
TSDBGrafana
Rich visualization layer with pre-built dashboards for cluster health, node performance, and workload metrics. Customizable and extensible for any use case.
VisualizationAlertmanager
Intelligent alert routing with de-duplication, grouping, and silencing. Route alerts to Slack, PagerDuty, Email, MS Teams, and custom webhooks.
AlertingNode Exporter
Deployed as a DaemonSet on every node, it exposes CPU, memory, disk I/O, and network metrics — giving you full visibility into host-level infrastructure.
InfrastructureKube-State-Metrics
Monitors Kubernetes API objects — deployments, pods, replica sets, services — tracking the desired vs. actual state of all your workloads.
K8s State Data Flow
How the Stack Works Together
A unified data pipeline from metric collection through intelligent alerting and rich visualization.
Orchestration
Prometheus Operator watches for ServiceMonitor and PodMonitor CRDs to auto-discover scrape targets.
Metric Collection
Node Exporter gathers hardware metrics; kube-state-metrics captures Kubernetes object states from the API server.
Scraping & Storage
Prometheus pulls metrics from all discovered endpoints and stores them as time-series data with configurable retention.
Rule Evaluation
PrometheusRule objects define alerting conditions. Triggered alerts are forwarded to Alertmanager for routing.
Alerting & Notification
Alertmanager de-duplicates, groups, and routes alerts to Slack, PagerDuty, email, or any webhook receiver.
Visualization
Grafana queries Prometheus to render real-time dashboards — pre-built for cluster health and fully customizable.
Quick Start
Deploy in Three Commands
From zero to full-stack monitoring in under five minutes using the official Helm chart from the Prometheus community.
terminal — bash
# Add the Prometheus community Helm repository $ helm repo add prometheus-community \ https://prometheus-community.github.io/helm-charts $ helm repo update # Create a dedicated monitoring namespace $ kubectl create namespace monitoring # Install the full kube-prometheus-stack $ helm install prometheus-stack \ prometheus-community/kube-prometheus-stack \ --namespace monitoring \ --values values.yaml
Production Best Practices
Built for Enterprise Scale
Follow battle-tested patterns to run a reliable, performant, and secure monitoring stack in production.
Persistent Storage
Configure Persistent Volumes for Prometheus and Grafana to survive pod restarts without losing metrics or dashboards.
Cardinality Control
Monitor and manage time-series cardinality to prevent memory explosions. Avoid high-cardinality labels like unique IDs.
High Availability
Run multiple Prometheus replicas with pod anti-affinity for zero-downtime monitoring across failure domains.
Long-Term Storage
Integrate with Thanos, Cortex, or Grafana Cloud via remote_write for historical data retention beyond 30 days.
Security & RBAC
Enforce network policies, OIDC/OAuth authentication for Grafana, and strict Kubernetes RBAC for the monitoring namespace.
ServiceMonitor Discovery
Use ServiceMonitor and PodMonitor CRDs for automatic, service-based metric target discovery — no manual config needed.
Resource Governance
Set CPU/memory requests and limits for every component to prevent resource starvation and OOM kills in production.
Full Observability Pillars
Extend beyond metrics: add Loki for logs and Tempo + OpenTelemetry for distributed tracing in a unified Grafana stack.
FAQ
Frequently Asked Questions
Everything you need to know about deploying and managing the kube-prometheus-stack.
It is a comprehensive Helm chart that deploys a full Kubernetes monitoring and alerting stack — including Prometheus, Grafana, Alertmanager, Node Exporter, kube-state-metrics, and the Prometheus Operator. It provides production-ready observability out of the box.
Standalone Prometheus requires manual configuration of scrape targets, alerting rules, and dashboards. The kube-prometheus-stack automates all of this using the Prometheus Operator pattern with CRDs like ServiceMonitor and PrometheusRule, plus it includes pre-configured Grafana dashboards and Alertmanager routing.
For a small cluster (1–5 nodes), Prometheus typically needs 2 CPU cores and 4–8 GB RAM. Grafana requires around 0.5 CPU and 512 MB RAM. For production clusters, scale resources based on the number of time series and scrape frequency. Always configure persistent volumes for data retention.
Absolutely. Custom Grafana dashboards can be provisioned via ConfigMaps or the Grafana UI. Custom alerting rules are defined using PrometheusRule CRDs, which the Operator automatically syncs with the Prometheus configuration. The stack is fully extensible.
Local Prometheus storage is recommended for 15–30 days of retention. For longer-term storage, configure remote_write to send metrics to solutions like Thanos, Cortex, Grafana Mimir, or managed services like Amazon Managed Prometheus and Grafana Cloud.
The default configuration is an excellent starting point, but production deployments should customize the values.yaml to enable persistent storage, set resource limits, configure HA replicas, define alert routing destinations (Slack/PagerDuty), and apply network policies for security.
You need a running Kubernetes cluster (v1.19+), Helm v3.x installed, and kubectl configured with cluster access. Ensure your cluster has sufficient resources — at minimum 2 vCPU and 4 GB RAM for the monitoring namespace. A default StorageClass is recommended for persistent volumes.
kube-prometheus-stack supports Kubernetes v1.19 and above, including all current EKS, GKE, and AKS managed versions. Helm v3.2+ is required. The chart is regularly tested against the latest Kubernetes releases and updated within days of new minor versions.
Run: helm repo update && helm upgrade --reuse-values monitoring prometheus-community/kube-prometheus-stack -n monitoring. Always review the chart changelog before upgrading, as CRD changes may require manual steps. Back up your Grafana dashboards and Prometheus data before major version upgrades.
CRDs are not managed by Helm on upgrades — you may need to update them manually for major chart versions. High-cardinality metrics can cause memory issues; always set cardinality limits. Prometheus does not natively support multi-tenancy — use Thanos or Cortex for multi-tenant setups. Windows nodes are not supported by Node Exporter.
Edit the alertmanager.config section in your values.yaml. Define receivers for Slack, PagerDuty, email, or webhook, then set up route rules to match alert labels to the correct receiver. Example: for Slack, provide the webhook URL and channel name under the slack_configs block of your receiver definition.