Existing coverage
The catalog includes application service health, both VPS hosts, backend API, PostgreSQL operations and failover, RabbitMQ queues and resources, worker pipeline SLOs, deployment ownership, backups, synthetic checks, logs, and usage quotas.
Production versus shadow
Worker deployment metrics distinguish active production ownership from shadow deployments. A stopped shadow service is not automatically a production outage. Use expected-active and deployment-identity signals when interpreting health.
Correlation
Start with the affected time range and service. Inspect request rates and error changes, then logs, queue depth, and deployment revision. Follow existing provider diagnostics for signals outside Grafana.
Instrumentation
Add bounded metrics or structured diagnostic fields only where a concrete gap prevents investigation. Preserve current behavior and the established NutsNews resource ownership.
Owner sign-in
Private dashboards use the portal Google session and preserve the NutsNews tab through authentication. Google OAuth changes apply to this observability portal, not the production workers or Qwen API.