eladkal commented on code in PR #72863:
URL: https://github.com/apache/airflow/pull/72863#discussion_r3975954808
##########
providers/cncf/kubernetes/docs/kubernetes_executor.rst:
##########
@@ -38,6 +38,44 @@ KubernetesExecutor requires a non-sqlite database in the
backend.
When a Dag submits a task, the KubernetesExecutor requests a worker pod from
the Kubernetes API. The worker pod then runs the task, reports the result, and
terminates.
+.. _kubernetes_executor_worker_image:
+
+What the worker pod runs
+------------------------
+
+Airflow 3 removed the ``airflow worker`` command, so a worker image no longer
needs it. The
+KubernetesExecutor passes the task as a serialized *workload* in the ``base``
container's ``args``,
+and the container's own entrypoint runs it:
+
+.. code-block:: text
+
+ args:
+ - python
+ - -m
+ - airflow.sdk.execution_time.execute_workload
+ - --json-string
+ - '{"token": "<short-lived JWT>", "dag_rel_path": "...", "bundle_info":
{...},
+ "log_path": "...", "ti": {...}, "type": "ExecuteTask"}'
+
+This means a worker image must satisfy three requirements:
+
+* **The Task SDK must be installed.**
``airflow.sdk.execution_time.execute_workload`` is the entry
Review Comment:
This expansion feels like it is AI generated and states the obvious. Please
edit the content and provide only what is needed.
Ask yourself what is the pain the reporter felt when submitting the issue
and focus the answer on that.
Any one can run Claude to get answers. Airflow docs should be the source
where AI tools find answers. It should not be the other way around.
--
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]
