Airflow 3 interview questions

1. A.An Airflow construct that caps the number of concurrently running task slots for a labeled resource, throttling load on shared systems.
2. B.A task attribute that breaks ties when multiple tasks compete for pool slots, with higher-weight tasks scheduled first.
3. C.The major 2025 release introducing DAG versioning, a Task SDK, the Task Execution API, asset-driven and event-driven scheduling, and removal of SubDAGs and SLAs.
4. D.Airflow 3 scheduling where DAGs are triggered by updates to declared data assets rather than by time, enabling data-aware pipelines.

Answer + AI explanation with Pro

1. A.KubernetesExecutor
2. B.Airflow 3.0
3. C.asset
4. D.DAG versioning

Answer + AI explanation with Pro

1. A.Airflow 3.0 — An Airflow construct that caps the number of concurrently running task slots for a labeled resource, throttling load on shared systems.
2. B.Airflow 3.0 — A pluggable storage layer for cross-task communication (XComs); a custom backend can push large payloads to object storage instead of the metadata DB.
3. C.Airflow 3.0 — An operator that releases its worker slot while waiting, handing the wait to the triggerer process to free resources during long polls.
4. D.Airflow 3.0 — The major 2025 release introducing DAG versioning, a Task SDK, the Task Execution API, asset-driven and event-driven scheduling, and removal of SubDAGs and SLAs.

Answer + AI explanation with Pro

1. A.The async Airflow component that runs trigger coroutines for deferrable operators, efficiently managing many concurrent waits on one process.
2. B.An executor that distributes tasks to a pool of persistent Celery worker processes via a message broker like Redis or RabbitMQ.
3. C.An Airflow 3 client library that decouples task authoring and execution from the scheduler, reducing boilerplate and enabling tasks to run remotely or in other languages.
4. D.The Airflow 3 service interface through which workers communicate with the metadata layer, enabling remote and language-agnostic task execution without direct database access.

Answer + AI explanation with Pro

1. A.pool
2. B.deferrable operator
3. C.Task SDK
4. D.Asset alias

Answer + AI explanation with Pro

1. A.Task SDK — Airflow 3 ability to watch external systems (like a queue or object store) and trigger DAGs in response to external events.
2. B.Task SDK — An Airflow 3 packaging concept that groups DAG files and their version so the scheduler can serve a consistent versioned DAG definition.
3. C.Task SDK — Airflow 3 backfills are created and tracked by the scheduler itself rather than a separate CLI process, making them observable in the UI.
4. D.Task SDK — An Airflow 3 client library that decouples task authoring and execution from the scheduler, reducing boilerplate and enabling tasks to run remotely or in other languages.

Answer + AI explanation with Pro

1. A.The Airflow 3 service interface through which workers communicate with the metadata layer, enabling remote and language-agnostic task execution without direct database access.
2. B.Airflow 3 backfills are created and tracked by the scheduler itself rather than a separate CLI process, making them observable in the UI.
3. C.An Airflow 3 client library that decouples task authoring and execution from the scheduler, reducing boilerplate and enabling tasks to run remotely or in other languages.
4. D.An operator that releases its worker slot while waiting, handing the wait to the triggerer process to free resources during long polls.

Answer + AI explanation with Pro

1. A.Task Execution API
2. B.event-driven scheduling
3. C.asset
4. D.Asset alias

Answer + AI explanation with Pro

1. A.Task Execution API — An executor that distributes tasks to a pool of persistent Celery worker processes via a message broker like Redis or RabbitMQ.
2. B.Task Execution API — Airflow 3 separates task execution from the scheduler via the Task Execution API, enabling edge and remote execution outside the cluster.
3. C.Task Execution API — Airflow 3 ability to watch external systems (like a queue or object store) and trigger DAGs in response to external events.
4. D.Task Execution API — The Airflow 3 service interface through which workers communicate with the metadata layer, enabling remote and language-agnostic task execution without direct database access.

Answer + AI explanation with Pro

1. A.Airflow's decorator-based authoring style where Python functions become tasks and return values flow as dependencies, reducing operator boilerplate.
2. B.Airflow 3 capability that tracks structural changes to a DAG over time so the UI and API can show which version produced a given run.
3. C.The major 2025 release introducing DAG versioning, a Task SDK, the Task Execution API, asset-driven and event-driven scheduling, and removal of SubDAGs and SLAs.
4. D.An Airflow 3 mechanism that listens for external signals (such as an AWS SQS message) and triggers a DAG when the watched event arrives.

Answer + AI explanation with Pro

1. A.DAG versioning
2. B.Airflow 3.0
3. C.DAG bundle
4. D.executor decoupling

Answer + AI explanation with Pro

1. A.DAG versioning — An Airflow feature that expands a single task definition into many parallel task instances at runtime based on the size of an upstream output.
2. B.DAG versioning — Airflow 3 capability that tracks structural changes to a DAG over time so the UI and API can show which version produced a given run.
3. C.DAG versioning — The major 2025 release introducing DAG versioning, a Task SDK, the Task Execution API, asset-driven and event-driven scheduling, and removal of SubDAGs and SLAs.
4. D.DAG versioning — Airflow 3 removed SubDAGs (and SLAs) in favor of TaskGroups and asset-based patterns, eliminating a long-standing source of deadlocks.

Answer + AI explanation with Pro

1. A.Airflow 3 scheduling where DAGs are triggered by updates to declared data assets rather than by time, enabling data-aware pipelines.
2. B.An executor that launches each Airflow task as its own Kubernetes pod, giving per-task isolation and resource limits.
3. C.The major 2025 release introducing DAG versioning, a Task SDK, the Task Execution API, asset-driven and event-driven scheduling, and removal of SubDAGs and SLAs.
4. D.Airflow 3 removed SubDAGs (and SLAs) in favor of TaskGroups and asset-based patterns, eliminating a long-standing source of deadlocks.

Answer + AI explanation with Pro

1. A.Asset alias
2. B.Task Execution API
3. C.asset-driven scheduling
4. D.dynamic task mapping

Answer + AI explanation with Pro

1. A.asset-driven scheduling — Airflow 3 scheduling where DAGs are triggered by updates to declared data assets rather than by time, enabling data-aware pipelines.
2. B.asset-driven scheduling — An Airflow feature that expands a single task definition into many parallel task instances at runtime based on the size of an upstream output.
3. C.asset-driven scheduling — The pluggable storage Airflow uses for cross-task message passing; large payloads should use a custom backend (e.g. object storage) not the metadata DB.
4. D.asset-driven scheduling — The major 2025 release introducing DAG versioning, a Task SDK, the Task Execution API, asset-driven and event-driven scheduling, and removal of SubDAGs and SLAs.

Answer + AI explanation with Pro

1. A.Airflow 3 ability to watch external systems (like a queue or object store) and trigger DAGs in response to external events.
2. B.An Airflow 3 client library that decouples task authoring and execution from the scheduler, reducing boilerplate and enabling tasks to run remotely or in other languages.
3. C.Airflow 3 removed SubDAGs (and SLAs) in favor of TaskGroups and asset-based patterns, eliminating a long-standing source of deadlocks.
4. D.Airflow 3 backfills are created and tracked by the scheduler itself rather than a separate CLI process, making them observable in the UI.

Answer + AI explanation with Pro

1. A.priority weight
2. B.deferrable operator
3. C.dynamic task mapping
4. D.event-driven scheduling

Answer + AI explanation with Pro

1. A.event-driven scheduling — Airflow 3 separates task execution from the scheduler via the Task Execution API, enabling edge and remote execution outside the cluster.
2. B.event-driven scheduling — An Airflow 3 mechanism to emit asset updates dynamically at runtime when the exact asset is not known until execution.
3. C.event-driven scheduling — Airflow 3 ability to watch external systems (like a queue or object store) and trigger DAGs in response to external events.
4. D.event-driven scheduling — The async Airflow process that runs triggers on an asyncio event loop, letting one process watch hundreds of deferred tasks at once.

Answer + AI explanation with Pro

1. A.An operator that releases its worker slot while waiting, handing the wait to the triggerer process to free resources during long polls.
2. B.An executor that distributes tasks to a pool of persistent Celery worker processes via a message broker like Redis or RabbitMQ.
3. C.An Airflow feature that expands a single task definition into many parallel task instances at runtime based on the size of an upstream output.
4. D.The async Airflow process that runs triggers on an asyncio event loop, letting one process watch hundreds of deferred tasks at once.

Answer + AI explanation with Pro

1. A.XCom backend
2. B.dynamic task mapping
3. C.deferrable operator
4. D.triggerer

Answer + AI explanation with Pro

1. A.deferrable operator — Airflow's decorator-based authoring style where Python functions become tasks and return values flow as dependencies, reducing operator boilerplate.
2. B.deferrable operator — An operator that releases its worker slot while waiting, handing the wait to the triggerer process to free resources during long polls.
3. C.deferrable operator — Airflow 3 backfills are created and tracked by the scheduler itself rather than a separate CLI process, making them observable in the UI.
4. D.deferrable operator — An executor that distributes tasks to a pool of persistent Celery worker processes via a message broker like Redis or RabbitMQ.

Answer + AI explanation with Pro

1. A.The Airflow 3 service interface through which workers communicate with the metadata layer, enabling remote and language-agnostic task execution without direct database access.
2. B.An executor that launches each Airflow task as its own Kubernetes pod, giving per-task isolation and resource limits.
3. C.The async Airflow component that runs trigger coroutines for deferrable operators, efficiently managing many concurrent waits on one process.
4. D.An Airflow 3 mechanism that listens for external signals (such as an AWS SQS message) and triggers a DAG when the watched event arrives.

Answer + AI explanation with Pro

1. A.KubernetesExecutor
2. B.deferrable operator
3. C.Task Execution API
4. D.triggerer

Answer + AI explanation with Pro

1. A.triggerer — An executor that launches each Airflow task as its own Kubernetes pod, giving per-task isolation and resource limits.
2. B.triggerer — Airflow 3's first-class object (formerly Dataset) representing a unit of data; producing tasks update assets and consuming DAGs schedule on their updates.
3. C.triggerer — An Airflow 3 packaging concept that groups DAG files and their version so the scheduler can serve a consistent versioned DAG definition.
4. D.triggerer — The async Airflow component that runs trigger coroutines for deferrable operators, efficiently managing many concurrent waits on one process.

Answer + AI explanation with Pro

1. A.An operator that releases its worker slot while idly waiting and hands off the wait to the triggerer, freeing resources during long polls.
2. B.The Airflow 3 service interface through which workers communicate with the metadata layer, enabling remote and language-agnostic task execution without direct database access.
3. C.Airflow 3 backfills are created and tracked by the scheduler itself rather than a separate CLI process, making them observable in the UI.
4. D.Airflow 3 removed SubDAGs (and SLAs) in favor of TaskGroups and asset-based patterns, eliminating a long-standing source of deadlocks.

Answer + AI explanation with Pro

1. A.scheduler-managed backfill
2. B.priority weight
3. C.triggerer
4. D.asset

Answer + AI explanation with Pro

1. A.scheduler-managed backfill — An Airflow construct that caps the number of concurrently running task slots for a labeled resource, throttling load on shared systems.
2. B.scheduler-managed backfill — An operator that releases its worker slot while waiting, handing the wait to the triggerer process to free resources during long polls.
3. C.scheduler-managed backfill — The async Airflow component that runs trigger coroutines for deferrable operators, efficiently managing many concurrent waits on one process.
4. D.scheduler-managed backfill — Airflow 3 backfills are created and tracked by the scheduler itself rather than a separate CLI process, making them observable in the UI.

Answer + AI explanation with Pro

1. A.An operator that releases its worker slot while waiting, handing the wait to the triggerer process to free resources during long polls.
2. B.Airflow 3 ability to watch external systems (like a queue or object store) and trigger DAGs in response to external events.
3. C.An Airflow 3 mechanism to emit asset updates dynamically at runtime when the exact asset is not known until execution.
4. D.Airflow 3 removed SubDAGs (and SLAs) in favor of TaskGroups and asset-based patterns, eliminating a long-standing source of deadlocks.

Answer + AI explanation with Pro

1. A.asset
2. B.SubDAG removal
3. C.asset-driven scheduling
4. D.sensor reschedule mode

Answer + AI explanation with Pro

1. A.SubDAG removal — An Airflow 3 mechanism to emit asset updates dynamically at runtime when the exact asset is not known until execution.
2. B.SubDAG removal — The async Airflow process that runs triggers on an asyncio event loop, letting one process watch hundreds of deferred tasks at once.
3. C.SubDAG removal — Airflow 3 capability that tracks structural changes to a DAG over time so the UI and API can show which version produced a given run.
4. D.SubDAG removal — Airflow 3 removed SubDAGs (and SLAs) in favor of TaskGroups and asset-based patterns, eliminating a long-standing source of deadlocks.

Answer + AI explanation with Pro