metrics-dashboard-generator

Metrics Dashboard Generator

Produce ready-to-import Prometheus scrape config and a Grafana dashboard JSON for a .NET service — built around the RED method (Rate, Errors, Duration) plus runtime health and your custom meters.

Input — point it at your project

Works on a target: a project or GitHub URL. Find which metrics exist: grep -rn "AddPrometheusExporter\|MapPrometheusScrapingEndpoint\|Meter(\|CreateCounter\|CreateHistogram\|AddOpenTelemetry" --include=*.cs <target>. Note custom meter/instrument names to add panels for them.

Prometheus scrape

scrape_configs:
  - job_name: 'orders-service'
    metrics_path: /metrics
    static_configs:
      - targets: ['orders-service:8080']

Service side: OpenTelemetry ... .WithMetrics(m => m.AddPrometheusExporter()) + app.MapPrometheusScrapingEndpoint();.

Grafana dashboard (panels generated)

• Rate — requests/sec: rate(http_server_request_duration_seconds_count[1m]) by route.
• Errors — 5xx ratio: sum(rate(...{http_response_status_code=~"5.."}[5m])) / sum(rate(...[5m])).
• Duration — p50/p95/p99: histogram_quantile(0.95, sum(rate(http_server_request_duration_seconds_bucket[5m])) by (le, route)).
• Runtime — GC collections, heap size, thread pool queue, working set (from AddRuntimeInstrumentation).
• Custom meters — a panel per custom counter/histogram you detected.

Output is a complete dashboard .json to import into Grafana, plus the scrape snippet.

Principles

• RED for request services, USE for resources — start there before vanity panels.
• Use histogram quantiles for latency, not averages (averages hide tail latency).
• Label by route/status, but watch cardinality — don't label by user id or raw path.
• Match metric names to the OTel/Prometheus conventions your exporter emits (verify the names).

How to use it & best prompts

"Generate a Grafana dashboard for my service", "Prometheus config + RED panels", "dashboard my custom order metrics", "p95 latency panel by route". Pairs with opentelemetry-setup.