Skip to content

studio-grafana

Studio Grafana — Every Self-Built Service Ships Its Observability There

Section titled “Studio Grafana — Every Self-Built Service Ships Its Observability There”
  1. One instance. Every self-built service or system ships its metrics, logs, dashboards and alerts to the studio Grafana — never a per-project Grafana, never a rival hosted service (Datadog, Honeycomb, a second Loki). Its live location and dashboard inventory are the SSOT doc docs/grafana-node-monitoring.md in The1Studio/arc-github-runners; never hardcode the URL elsewhere.
  2. Dashboards are files in The1Studio/grafana-deploy, never click-built. One ConfigMap per dashboard, shipped by that repo’s CI. The instance has no persistence: a UI-built object is unreviewed and gone on the next restart or deploy.
  3. Every signal gets a route. A new service registers its scrape target (ServiceMonitor in The1Studio/arc-github-runners k8s/monitoring/) AND ships a dashboard AND a PrometheusRule routed through Alertmanager to the studio Discord channel. A service nobody would be notified about is not instrumented.
  4. Secrets stay out of the repo. Webhook URLs, API tokens and the admin password come from the environment or a k8s Secret, per modules/t1k-base/rules/security.md.

How: the t1k-grafana skill is the procedure — instrument → register → dashboard → alert. Related: docker-volume-discipline.md (a NON-studio compose Grafana keeps grafana.db on a named volume) · managed-datastore-write-cost-discipline.md (a metrics pipeline is a write path; cardinality costs memory) · agents/t1k-devops-engineer.md.

Why one instance, why files over clicks, and where the instance lives: docs/studio-grafana.md.