Backstage operations
Diagnose from evidence before restarting or redeploying, and preserve clear ownership between the Backstage service and its infrastructure.
When to invoke
- "Backstage will not start or is not ready."
- "Catalog refreshes or scheduled tasks are failing."
- "Diagnose database, memory, or latency problems."
- "Create a production operations checklist."
Procedure
- Detect adopter, core development, Open Horizons, RHDH, or unknown mode and select the real runtime boundary.
- Capture deployment version, config selection, database client, recent changes, health status, resource limits, and error evidence.
- Inspect application logs and traces with sensitive values redacted.
- Classify the failure as configuration, identity, integration, database, scheduler, plugin, dependency, network, resource, or deployment.
- Reproduce safely with the narrowest health endpoint, package test, or local startup path.
- Fix the root cause in the owning package or configuration layer.
- Require approval before restarts, rollouts, database changes, traffic changes, or production mutations.
- Validate readiness, liveness, login, catalog, templates, TechDocs, critical plugins, and scheduled tasks after remediation.
- Record rollback, residual risk, and infrastructure handoffs.
Operational evidence
- Backstage and package versions.
- Effective environment and configuration files.
- Health endpoint result.
- Database connectivity and migration state without credentials.
- Recent error signatures and affected plugin.
- Resource saturation, scheduler backlog, provider throttling, or integration failures.
Output template
## Backstage operations report
**Environment:** <environment>
**Version:** <version>
**Impact:** <users and capabilities>
| Evidence | Finding | Action | Result |
| --- | --- | --- | --- |
### Approval-gated operations
- <operation and status>
### Rollback
- <verified rollback or blocker>
Quality gate
- Runtime mode, environment, version, and user impact are evidenced.
- Logs and telemetry are redacted and tied to a root-cause hypothesis.
- No restart or redeploy occurred before diagnosis and approval.
- Critical Backstage capabilities are revalidated after the fix.
- Database and integration mutations include rollback.
- Infrastructure-only issues are handed to the owning platform workflow.