How do I point Prometheus at the /metrics endpoint when running the monitoring stack?

[ATemova]

I've got rosscope running and the /metrics endpoint is exposed, but I'm not sure how to get the bundled Prometheus + Grafana stack to actually scrape it. Do I need to edit the scrape config, and what target should I use when everything's running via Docker Compose?

[ATemova]

Two things need to happen: bring the monitoring services up, and make sure Prometheus points at the right target.

1. The Prometheus + Grafana services are behind a Compose profile. A plain docker compose up only starts the core services plus the synthetic sim fleet — prometheus and grafana are gated behind profiles: ["monitoring"], so they're skipped unless you opt in:
   docker compose --profile monitoring up
   That starts Prometheus on :9090 and Grafana on :3000. Grafana has anonymous viewer access enabled, so you can open http://localhost:3000 directly without logging in.

2. You don't need to edit the scrape config — the target is the api service. The /metrics endpoint is served by the api service on port 8000, and inside the Compose network you address services by name, not localhost. The bundled monitoring/prometheus.yml already has this correct:
   `scrape_configs:

• job_name: rosscope
  metrics_path: /metrics
  static_configs:
  • targets: ['api:8000']`
    So the target is api:8000 with metrics_path: /metrics. The common mistake is switching this to localhost:8000 or 127.0.0.1:8000 — from inside the Prometheus container that loops back to Prometheus itself, not the API, and the target shows DOWN.
    If the target is still DOWN after starting the monitoring profile:

Check the api container is healthy. Prometheus depends_on: [api], but if the API is restarting (e.g. DB not ready yet) there's nothing to scrape.
Verify from the host: curl http://localhost:8000/metrics should return Prometheus-format text.
Look at Status → Targets in the Prometheus UI — a connection-refused there points to a name/port mismatch against api:8000.

Once the target flips to UP, the Grafana dashboards populate automatically, since they're provisioned against the Prometheus datasource.