Cloudera Docs

Prometheus metrics integration with Grafana

Kudu provides native integration with Prometheus by exposing metric endpoints on Master and Tablet Server daemons. This integration allows you to monitor historical performance, set up alerts, and create visual dashboards in Grafana.

Key capabilities

  • Prometheus metric endpoint — Exposes point-in-time cluster metrics at the /metrics_prometheus endpoint.

  • Automated HTTP service discovery — Discovers all active Master and Tablet Server instances dynamically at the /prometheus-sd endpoint on the leader Master, eliminating the need for static target lists.

  • Entity label support — Exports entity identity and attributes as Prometheus labels when you set the --prometheus_use_entity_labels configuration flag to true.

  • Host identification — Includes hostnames as metric labels to distinguish metrics across nodes. You can toggle this behavior by using the --metrics_prometheus_export_hostname configuration flag.

  • Targeted metric filtering — Accepts filter parameters such as level, types, ids, metrics, and attributes directly on the /metrics_prometheus endpoint to narrow scraped data at the source.

  • Authentication support — Supports Bearer token authentication for Kerberos or SPNEGO-secured clusters by configuring the --webserver_prometheus_token_cmd configuration flag.