kosiew commented on code in PR #64960:
URL: https://github.com/apache/airflow/pull/64960#discussion_r3295862848
##########
providers/databricks/src/airflow/providers/databricks/triggers/databricks.py:
##########
@@ -53,6 +54,7 @@ def __init__(
caller: str = "DatabricksExecutionTrigger",
) -> None:
super().__init__()
+ validate_deferrable_databricks_retry_args(retry_args, owner=caller)
Review Comment:
Yes, intentionally. For deferrable execution the trigger kwargs must be
serialization-safe as soon as the trigger is created, because the trigger is
passed through Airflow's trigger serialization boundary. If we allow the
trigger instance to be created with non-serializable `retry_args`, the failure
moves later into generic serialization code and the user loses the
Databricks-specific guidance.
This also matches the issue scope: fail fast with a clear `ValueError`, and
ensure the Databricks trigger does not accept unsupported callable retry args.
Constructor-time validation keeps invalid trigger state from existing at all.
There is precedent for trigger constructors rejecting invalid arguments,
e.g. `S3KeysUnchangedTrigger` validates negative `inactivity_period`, MWAA
triggers validate conflicting states, and `DateTimeTrigger` rejects naive
datetimes. So this is not meant as a broad gatekeeper pattern, only validation
for an argument that cannot safely cross the deferrable boundary.
--
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]
