Skip to main content

Open Grafana and investigate service metrics

Documentation for Tutela v3.9.0.

Open the Grafana dashboards supplied with your deployment to compare service health, latency, errors and resource use during a service disruption. Use Tutela's Activity and AI Spend views for individual request evidence and cost allocation.

Prerequisites​

  • Administrator access to Settings > Support & diagnostics.
  • Grafana included in your deployment and an authorized dashboard account or launch configuration.
  • The time range and service involved in the problem.

Open dashboards​

  1. Open Settings > Support & diagnostics and find the Grafana dashboards card.
  2. Review its health and launch status. Not included means this deployment does not provide Grafana through that feature. Unavailable means the check or launch could not establish access; it is not a healthy result.
  3. Select Open Grafana when available. Tutela creates a short-lived launch session for the configured dashboard address.
  4. Set Grafana's time range to the affected interval. Confirm the dashboard timezone before comparing it with a request timestamp.
  5. Select the affected service and compare latency, errors, request volume and resource use for that same interval. A chart without data cannot establish that the service is healthy.

Interpret the charts​

Use request duration together with request count. A small number of requests can produce an unstable percentile; rising latency with rising errors or sustained load deserves investigation.

Compare queue or processing delay with when Activity events appeared. Compare cache errors with inspection availability. Follow the relevant runbook before changing capacity or policy behavior.

Grafana contains operational measurements. It is not the chargeback ledger and should not expose prompts, user content or credentials. Use AI Spend for organizational cost allocation and Activity for the recorded outcome of an individual request.

If Grafana does not open​

  1. Record the card's request ID and visible error. Check Service health to determine whether only dashboard access or other services are affected.
  2. Ask the deployment operator to confirm the configured Grafana address, certificate, authentication and routing. A successful application sign-in does not independently establish Grafana access.
  3. For a Helm installation, compare the installed chart's approved ingress peers and OAuth destinations with its configuration. Keep the existing network restrictions; do not enable unrestricted access to make a dashboard load.
  4. For an AWS installation, have the operator inspect the load balancer target health and the dashboard service for the affected interval.
  5. Retry the launch after the specific cause is corrected. Confirm that the selected dashboard contains current measurements, not just that the browser opened.

Validate​

  • Grafana opened at the expected address using the intended access path.
  • The selected time range and service match the reported problem.
  • Measurements are current, and missing series are understood.
  • Any configuration change has an owner and a recorded previous value.

Next steps​