uranusjr commented on code in PR #31840: URL: https://github.com/apache/airflow/pull/31840#discussion_r1233592419
########## docs/apache-airflow/authoring-and-scheduling/deferring.rst: ########## @@ -166,3 +166,21 @@ Airflow tries to only run triggers in one place at once, and maintains a heartbe This means it's possible, but unlikely, for triggers to run in multiple places at once; this is designed into the Trigger contract, however, and entirely expected. Airflow will de-duplicate events fired when a trigger is running in multiple places simultaneously, so this process should be transparent to your Operators. Note that every extra ``triggerer`` you run will result in an extra persistent connection to your database. + + +Difference between Mode='reschedule' and Deferrable=True in Sensors +------------------------------------------------------------------- + +In Airflow, Sensors wait for specific conditions to be met before proceeding with downstream tasks. Sensors have two options for managing idle periods: mode='reschedule' and deferrable=True. + ++--------------------------------------------------------+--------------------------------------------------------+ +| Mode='reschedule' | Deferrable=True | ++========================================================+========================================================+ +| Continuously reschedules | Pauses execution when idle | ++--------------------------------------------------------+--------------------------------------------------------+ +| Resource Usage is Higher | Resource Usage is Lower | ++--------------------------------------------------------+--------------------------------------------------------+ +| Changing conditions over time | Waiting for external events | Review Comment: Also should worth a few more words. The main difference here is, a sensor is a running process (worker) that continuously sniffs an external state and change its own state sccordingly. A deferred task, however, is _stopped_ to wait until the external state changes, and relaunched with the new state. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
