Why are my metrics missing in Grafana?

Work through these checks before contacting Support:

Wrong data source for your cluster’s region. The most common cause. Verify the dashboard’s data source matches the Super Region where your cluster runs. See Data sources for the region-to-endpoint mapping.
Missing IAM role. Your user needs the Observability Viewer role to see any dashboards, plus Billing Viewer for the Usage by Product and Zone dashboard.
Invalid API token (direct API or self-hosted Grafana). Every request must include a valid Authorization: Bearer [ACCESS-TOKEN] header. Test it:
```
curl -G "https://observe.coreweave.com/api/v1/query" \
  -H "Authorization: Bearer [ACCESS-TOKEN]" \
  --data-urlencode "query=up"
```
If it fails, generate a new API Access Token.
No Prometheus Operator on the cluster. Clusters created after July 7, 2025 do not ship with it, so custom application metrics are not collected. Install it and add a PrometheusAgent with remoteWrite. See Forward application metrics.
Billing metric inflated by control plane Nodes. Clusters created before July 7, 2025 include a non-billable cpu-control-plane Node Pool. Filter with billing:instance:total{node_pool_name!="cpu-control-plane"}. See Usage monitoring.
Log expired. Some log classes expire after two weeks. If a log is missing from your Super Region, check the legacy endpoint observe-legacy.coreweave.com.