Check Health and Monitoring
Use separate Woodpecker publication and Watchtower reconciliation evidence for image identity. A health handler does not expose registry tag identity. Use the health response and monitoring timestamp only for availability evidence. An anonymous Tracker or Nexus request redirects through Caddy to Authentik and cannot prove backend health. Do not include cookies, tokens, or private response bodies.
Prerequisites
Section titled “Prerequisites”Record the expected immutable image identity from Woodpecker. Record the Watchtower reconciliation time from the production operator. Obtain monitoring and production container health access.
Actions
Section titled “Actions”- Check the Website at
/. - Check Learn at
/api/health. - Inspect Tracker’s internal
/api/healthcontainer healthcheck. - Inspect Nexus’s internal
/api/healthcontainer healthcheck. - Check pckg at
/health/ready. - Verify the signed-in Tracker and Nexus routes through Authentik.
- Compare the public and internal check times with Watchtower logs.
- Inspect the same reconciliation window at
monitor.beskid-lang.org.
| Boundary | Healthy evidence | Next diagnostic |
|---|---|---|
| Website | Successful HTTP status for /. | Site container and proxy logs. |
| Learn | Successful HTTP status for /api/health. | One safe lesson check and runtime-kit evidence. |
| Tracker | Healthy internal /api/health container check and signed-in route. | SQLite volume and Authentik forward-auth configuration. |
| Nexus | Healthy internal /api/health container check and signed-in route. | Proxy trust boundary and nexus-data. |
| pckg | Successful HTTP status for /health/ready. | PostgreSQL connectivity and artifact volume. |
Expected result
Section titled “Expected result”Public Website, Learn, and pckg routes respond. Tracker and Nexus report healthy internal checks and load after Authentik sign-in. Container and public evidence agree on the deployment window. Separate release and deployment evidence identifies the expected images.
Recovery
Section titled “Recovery”If one boundary fails, preserve its correlation evidence and inspect only that service. If several services fail, inspect the proxy and shared lane first. Ask the production operator to restore the previous state when the deployment caused the failure. Repeat all checks after the operator completes recovery.