uranusjr commented on code in PR #32000:
URL: https://github.com/apache/airflow/pull/32000#discussion_r1264387657
##########
airflow/models/dag.py:
##########
@@ -2730,22 +2730,50 @@ def add_logger_if_needed(ti: TaskInstance):
)
tasks = self.task_dict
+ # Special case when retry is set and DAG is paused
+ task_try_number: dict[str, int] = {}
+
self.log.debug("starting dagrun")
# Instead of starting a scheduler, we run the minimal loop possible to
check
# for task readiness and dependency management. This is notably faster
# than creating a BackfillJob and allows us to surface logs to the user
while dr.state == State.RUNNING:
schedulable_tis, _ = dr.update_state(session=session)
- try:
- for ti in schedulable_tis:
+
+ for ti in schedulable_tis:
+ try:
add_logger_if_needed(ti)
ti.task = tasks[ti.task_id]
_run_task(ti, session=session)
- except Exception:
- self.log.info(
- "Task failed. DAG will continue to run until finished and
be marked as failed.",
- exc_info=True,
- )
+ except Exception:
+ if ti.state == TaskInstanceState.UP_FOR_RETRY:
+ try_number =
task_try_number.get(f"{ti.task_id}_{ti.map_index}", 0)
+ if try_number > ti.max_tries:
+ ti.set_state(TaskInstanceState.FAILED)
+ else:
+ task_try_number[f"{ti.task_id}_{ti.map_index}"] =
try_number + 1
+ self.log.info(
+ "Task failed. DAG will continue to run until finished
and be marked as failed.",
+ exc_info=True,
+ )
+ for ti in dr.get_task_instances(session=session):
Review Comment:
Hmm, this sounds problematic to me. Removing `session` means the function
opens a new database session, which would not contain changes made in the
session in the outer scope. If this causes state issue we should probably try
to figure out what the cause is.
--
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]
