diogosilva30 commented on code in PR #65943:
URL: https://github.com/apache/airflow/pull/65943#discussion_r3209074235
##########
providers/edge3/src/airflow/providers/edge3/cli/worker.py:
##########
@@ -447,17 +451,77 @@ def _run_job_via_supervisor(self, workload: ExecuteTask,
results_queue: Queue) -
results_queue.put(e)
return 1
- def _launch_job(self, workload: ExecuteTask) -> tuple[Process,
Queue[Exception]]:
+ def _launch_job_subprocess(self, workload: ExecuteTask) ->
subprocess.Popen:
+ """Launch workload via a fresh Python interpreter
(subprocess.Popen)."""
+ env = os.environ.copy()
+ if self._execution_api_server_url:
+ env["AIRFLOW__CORE__EXECUTION_API_SERVER_URL"] =
self._execution_api_server_url
+
+ # Keep stderr off a PIPE: the worker only inspects stderr after the
task finishes,
+ # so a verbose child could otherwise fill the pipe buffer and block
forever.
+ with tempfile.NamedTemporaryFile(
+ prefix="airflow-edge-task-stderr-", suffix=".log", delete=False
+ ) as stderr_file:
+ stderr_file_path = Path(stderr_file.name)
+ try:
+ process = subprocess.Popen(
+ [
+ sys.executable,
+ "-m",
+ "airflow.sdk.execution_time.execute_workload",
+ "--json-string",
+ workload.model_dump_json(),
+ ],
+ env=env,
+ start_new_session=True,
+ stderr=stderr_file,
Review Comment:
The Queue approach works in the fork path because the child inherits the
multiprocessing state, including the Queue itself.
With `subprocess.Popen(...)` we start a completely fresh Python interpreter,
so there is no shared Queue unless we build a separate IPC layer
(pipe/socket/fd passing/etc).
We could do that, but it adds quite a bit of complexity compared to the
current tempfile approach. The tempfile also avoids PIPE deadlocks and still
captures early bootstrap/import failures before any IPC channel would be
initialized.
--
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]
