Skip to content

Troubleshoot Healthcheck Core

Remove it from liveness; register it for readiness. Kubernetes restarts on failed liveness but stops routing on failed readiness.

Increase its timeoutMs only after checking latency. Ensure the check observes ctx.signal; abandoned work can otherwise continue after the report completes.

Inspect summary and each result’s critical flag. A non-critical degraded result is still visible even when the aggregate remains healthy.

Exporter packages read report.metrics, populated by collectors. Metrics nested inside check results remain part of JSON check output.

Disable details and enable redaction. Never put tokens, connection strings, user identifiers, or raw response bodies into metrics, labels, or details.