Skip to content

Dashboards & Visualisation (Grafana)

Status: 📋 Planned — Grafana deployment pending; Prometheus data is already being collected.

Prometheus metrics are live and ready to be visualized. No dashboards are deployed today.

See Metrics reference for the full list of available metrics.

Intended dashboards

Service Overview

  • request rate (soccer_requests_total),
  • latency percentiles (soccer_request_duration_seconds),
  • error rate (soccer_errors_total),
  • active model version (soccer_model_version).

Async Processing

  • Celery queue depth (soccer_celery_queue_length),
  • active workers (soccer_celery_workers_active),
  • retry and backlog trends.

Infrastructure

  • Kubernetes CPU / memory / pod restarts (requires pg_exporter and k8s metrics — not yet configured).

Design principles

  • Focus on trends, not single points.
  • Align panels with SLOs.
  • Annotate deployments and model promotions.
  • Avoid excessive label cardinality.