jason810496 commented on code in PR #75387:
URL: https://github.com/apache/airflow/pull/75387#discussion_r4397600420
##########
airflow-core/docs/core-concepts/tasks.rst:
##########
@@ -145,6 +145,15 @@ if the task has not stopped within ``[core]
killed_task_cleanup_time`` seconds a
(for example because it is blocked in native code), the supervisor sends it
SIGTERM, and SIGKILL after the same
period again. Tasks run by a language SDK do not raise ``AirflowTaskTimeout``
and are stopped only this way.
+A task that does not set ``execution_timeout`` itself gets ``[core]
default_task_execution_timeout``, which is read in
+different places depending on how the task runs:
+
+- A Python task reads it on the worker, which parses the Dag file again before
running the task.
+- A task run by a language SDK reads it where the Dag processor runs, when the
Dag is parsed and serialized.
+
+Set ``[core] default_task_execution_timeout`` to the same value on the Dag
processor and the workers, so both kinds
+of task get the same default.
+
Review Comment:
Let's remove this part, since this part should be fixed after the new plan I
proposed above.
##########
airflow-core/src/airflow/api_fastapi/execution_api/datamodels/taskinstance.py:
##########
@@ -559,7 +559,9 @@ class TIRunContext(BaseModel):
"""
The task's ``execution_timeout`` in seconds, taken from its serialized Dag.
- The supervisor enforces the timeout with this value. ``None`` when no
timeout is set.
+ The supervisor enforces the timeout of a task run by a language SDK with
this value. A Python task
+ enforces its own ``execution_timeout`` and sends that to the supervisor
instead, since the server
Review Comment:
We should revert this part from the API server in this PR. The root cause
was mentioned in https://github.com/apache/airflow/issues/75791 and
https://github.com/apache/airflow/pull/75795/ just merged.
https://github.com/apache/airflow/blob/840d3a41c633abb1d5d97a8791af066b14b2a069/airflow-core/src/airflow/api_fastapi/execution_api/routes/task_instances.py#L407-L408
--
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]
