yuqian90 commented on pull request #12312:
URL: https://github.com/apache/airflow/pull/12312#issuecomment-726535906


   That being said, I wasn't super aware of the feature of dynamically 
generating unique task_id when `@task` decorator is used. Now looking at the 
code a bit more, I have some concerns with using the dynamic task_id feature 
too much in production DAGs which may evolve incrementally. 
   
   For example, if there are many `@task` decorated tasks without explicitly 
given `task_id`. Their task_ids will be generated sequentially: task__1, 
task__2, task__3, etc. After the DAG goes into production, one day someone 
inserts a new task before `task__2`. The task_ids after that will all be 
shifted forward by one place. This is going to produce `task__1, task__2, 
task__3, task__4`. But at this point the task__3 is no longer the same 
`task__3` as before. This is going to cause confusion when looking at 
historical DagRuns. E.g. history, log files, etc.
   
   I think we should make it clear somewhere that the dynamic task_id feature 
should be used carefully if the DAG is expected to change in the future.


----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to