Skip to content

Runbooks

Each runbook below is written against a specific alert or failure condition: verification commands first, then causes and mitigations. See troubleshooting.md for a symptom-first index instead, and conditions.md for the full condition type/reason registry these runbooks reference.

RunbookAlert / condition
Operator DownFleetOperatorDown — no healthy operator instances for 5 minutes
High Reconcile Error RateFleetReconcileErrorRateHigh — reconcile errors > 0.1/s for 10 minutes
Rate Limit SaturationFleetRateLimitSaturation — p95 rate-limiter wait > 0.9s for 5 minutes
Webhook UnavailableFleetWebhookRejectionRateHigh — admission webhook returning 500s for 5 minutes
Finalizer StuckA CR has deletionTimestamp set but its finalizer has not been removed after 5+ minutes
Pipeline Name Scope MigrationOperational guide for enabling --pipeline-name-scope=namespace

The alert names above match the PrometheusRule shipped by the Helm chart when alerts.enabled: true.