Airflow Mastery (Senior → Principal)

Operate

• Start from orchestration semantics, data dependencies, and failure ownership.
• Treat Airflow as workflow control infrastructure, not as a place to hide business logic.
• Prefer DAGs that remain readable, recoverable, and operationally predictable.
• Optimize for reliability, clear retries, and low scheduler toil.

Default Standards

• DAG boundaries should reflect real workflow ownership.
• Tasks should be idempotent and retry-aware.
• Scheduling semantics must match data availability and SLA reality.
• Platform governance should control DAG sprawl and dependency chaos.
• Airflow should orchestrate systems, not become the system.

References

• DAG architecture and task boundary design: references/dag-architecture-and-task-boundary-design.md
• Scheduling, SLAs, and dependency semantics: references/scheduling-slas-and-dependency-semantics.md
• Executors, workers, and platform scaling: references/executors-workers-and-platform-scaling.md
• Retries, idempotency, and external side effects: references/retries-idempotency-and-external-side-effects.md
• Governance, multi-team DAG ecosystems, and ownership: references/governance-multi-team-dag-ecosystems-and-ownership.md
• Observability and debugging: references/observability-and-debugging.md
• Reliability and operations: references/reliability-and-operations.md
• Incident runbooks: references/incident-runbooks.md