Principle:Risingwavelabs Risingwave Metrics Dashboard Generation

Knowledge Sources	Grafana Documentation RisingWave
Domains	Observability, Monitoring, Code_Generation
Last Updated	2026-02-09 07:00 GMT

Overview

A code generation approach that programmatically constructs Grafana monitoring dashboards from Python definitions, enabling version-controlled, parameterized observability for streaming database clusters.

Description

Metrics Dashboard Generation takes a dashboards-as-code approach to observability. Instead of manually creating Grafana dashboards through the UI, RisingWave defines dashboards programmatically in Python using the grafanalib library. These Python definitions are compiled to Grafana JSON and provisioned automatically.

Two dashboard profiles are generated:

• User Dashboard: Simplified operational view with overview, CPU, memory, network, storage, streaming, and batch panels
• Dev Dashboard: Detailed internal metrics with 40+ sections covering Hummock compaction, streaming operators, barrier latency, CDC throughput, and Iceberg operations

The dashboards are parameterized with Grafana template variables for namespace, cluster name, and datasource, making them reusable across environments.

Usage

Use dashboard generation when:

• Setting up monitoring for a new RisingWave deployment
• Customizing dashboards for specific operational needs
• Adding new metrics panels for new features
• Maintaining dashboard definitions in version control

Theoretical Basis

Dashboard-as-Code Pipeline:
    Python definitions (grafanalib)
        |
        v
    generate.sh (grafanalib → JSON)
        |
        v
    Grafana provisioning (auto-load JSON)
        |
        v
    Grafana UI (rendered dashboards)

Template Variables:
    $namespace → Kubernetes namespace filter
    $cluster → Cluster name filter
    $datasource → Prometheus data source
    $node → Individual node filter

Related Pages

Implemented By

• Implementation:Risingwavelabs_Risingwave_Grafana_Dashboard_Generation