Pipeline On-Call Runbook Builder

Write an on-call runbook for a tier-1 data pipeline that a capable engineer can use at 3 a.m. without tribal knowledge. Pipeline purpose and business impact: [description] Architecture and dependencies: [components] SLOs and critical delivery windows: [targets] Dashboards, logs, and lineage tools: [links or names] Common failure modes: [known incidents] Safe operational commands or procedures: [details] Escalation contacts and ownership: [roles, not personal secrets] Create: 1. Fast impact assessment and severity classification 2. Decision tree from symptom to likely component 3. Checks for freshness, volume, schema, quality, infrastructure, and upstream outages 4. Exact evidence to collect before retrying or mutating state 5. Safe retry, replay, pause, failover, and rollback procedures 6. Guardrails that prevent duplicate writes or destructive backfills 7. Communication templates for stakeholders and incident channels 8. Escalation triggers based on time, impact, and uncertainty 9. Recovery validation and criteria for declaring resolution 10. Post-incident follow-up checklist and runbook ownership Separate read-only diagnostics from state-changing actions. Mark dangerous operations with prerequisites, blast radius, and rollback steps.