Re: [I] Improve Airflow core docs regarding Dag versioning (AIP-63) [airflow]

[via GitHub]

MonsterChenzhuo commented on issue #61344:
URL: https://github.com/apache/airflow/issues/61344#issuecomment-4051954493

   > > <img alt="图像" width="733" height="449" 
src="https://private-user-images.githubusercontent.com/60029759/562047580-9e974dda-2a7f-478f-867a-fe640a5d7bfa.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.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.7woYZUdYo2dRUtzP2wHHaKZa6NMCd0iUL29Oja1daAA";>
   > > [@shahar1](https://github.com/shahar1)  
[@rahulmansharamani14](https://github.com/rahulmansharamani14)我有一个关于DAG逻辑的技术问题。今天是2026年3月12日，我刚刚向一个现有的DAG中添加了一个名为test1的新运算符。
   > > 我希望回填/运行特定运算符，回填时间为 2026 年 3 月 1 日至 2026 年 3 月 11 
日。但是，我找不到在运算符级别触发此操作的按钮。在网格/图表视图中，此新任务的“方格”在这些过去的日期是空白的，并且无法点击。
   > > 
因此，我不得不重新填充那几天的整个DAG，这效率非常低。是不是我在用户界面中遗漏了什么？或者有没有更好的方法，可以在不重新运行所有内容的情况下，针对历史数据范围触发单个任务？
   > > 补充：我已阅读您发送的文章，但我仍然不太理解。
   > 
   > 据我所知，如果某个任务在原始的 DAG 运行中不存在，Airflow 不提供追溯运行历史日期单个任务的选项。
   > 
   > 虽然现在可以针对过去的逻辑日期运行当前版本的 DAG，但 Airflow 中的任务始终在 DAG 
运行的上下文中执行。如果独立地针对过去的日期运行任务，则意味着在没有完整 DAG 上下文（包括定义该次运行的依赖关系和结构）的情况下执行该任务。
   > 
   > 由于对 DAG 的更改（例如添加任务或修改依赖关系）实际上会改变其上下文，因此，如果不重新创建相应的 DAG 
运行，则回溯执行单个任务将无法反映正确的工作流状态。因此，通常的做法是触发或回填相关日期的 DAG，而不是单独运行单个任务。
   
   In my view, version traceability is not always a mandatory requirement. For 
scenarios where it's unnecessary, the current overhead of tracing the entire 
DAG history feels excessive.
   
   I’m considering whether we could introduce a configuration toggle for users 
who don't require strict versioning. Here is the logic:
   
   Current Behavior: When a new task (e.g., test1) is added, Airflow generates 
a new DagVersion and writes the serialized DAG. However, historical DagRuns 
(e.g., from 2026-03-01 to 03-11) do not have corresponding TaskInstance rows 
for test1, causing 'blank and unclickable' cells in the Grid view.
   
   Proposed Improvement: Upon DAG parsing, if this toggle is enabled, Airflow 
would automatically backfill the missing TaskInstance rows for historical 
DagRuns. This ensures the Grid view represents test1 with a 'No Status/Not Run' 
state instead of a gap, providing a more consistent UI experience."
   
   > > <img alt="Image" width="733" height="449" 
src="https://private-user-images.githubusercontent.com/60029759/562047580-9e974dda-2a7f-478f-867a-fe640a5d7bfa.png?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3NzMzNjk0NDcsIm5iZiI6MTc3MzM2OTE0NywicGF0aCI6Ii82MDAyOTc1OS81NjIwNDc1ODAtOWU5NzRkZGEtMmE3Zi00NzhmLTg2N2EtZmU2NDBhNWQ3YmZhLnBuZz9YLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUFWQ09EWUxTQTUzUFFLNFpBJTJGMjAyNjAzMTMlMkZ1cy1lYXN0LTElMkZzMyUyRmF3czRfcmVxdWVzdCZYLUFtei1EYXRlPTIwMjYwMzEzVDAyMzIyN1omWC1BbXotRXhwaXJlcz0zMDAmWC1BbXotU2lnbmF0dXJlPWYxNTU0YzdhZGE0ZWQ0NzE5Y2ZmYTA5ZWM0NzQ2YjVmNTAxN2MyOGYxNDhlMjY2YTkzYmE1NDE1MzE4OGFkMmEmWC1BbXotU2lnbmVkSGVhZGVycz1ob3N0In0.7kQKqR6T9xPmEL6VMm5gS-WhW4_wBp4QQeqvhRYPR5k";>
   > > [@shahar1](https://github.com/shahar1)  
[@rahulmansharamani14](https://github.com/rahulmansharamani14) I have a 
technical question regarding DAG logic. Today is March 12, 2026, and I just 
added a new Operator named test1 to an existing DAG.
   > > I want to backfill/run only this specific Operator for the historical 
period of March 1, 2026, to March 11, 2026. However, I cannot find a button to 
trigger this at the Operator level. In the Grid/Graph view, the 'squares' for 
this new task are empty for those past dates and are not clickable.
   > > As a result, I am forced to backfill the entire DAG for those days, 
which is very inefficient. Am I missing something in the UI, or is there a 
better way to trigger a single task for a historical range without re-running 
everything?
   > > To add: I have read the article you sent, but I still don't understand 
it very well.
   > 
   > AFAIK, Airflow does not provide an option to retrospectively run a single 
task for historical dates if that task did not exist in the original Dag run.
   > 
   > While it is now possible to run the current version of a Dag for past 
logical dates, tasks in Airflow are always executed within the context of a Dag 
run. Running a task independently for past dates would mean executing it 
without the full Dag context, including the dependencies and structure that 
defined that run.
   > 
   > Since changes to the Dag (such as adding tasks or modifying dependencies) 
effectively change that context, executing a single task retrospectively 
without recreating the corresponding Dag runs would not reflect the correct 
workflow state. For that reason, the typical approach is to trigger or backfill 
the Dag for the relevant dates rather than running an individual task in 
isolation.
   
   


-- 
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: commits-unsubscr...@airflow.apache.org

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org