jason810496 commented on code in PR #60810:
URL: https://github.com/apache/airflow/pull/60810#discussion_r2812719704
##########
airflow-core/src/airflow/api_fastapi/execution_api/datamodels/taskinstance.py:
##########
@@ -304,6 +305,40 @@ class DagRun(StrictBaseModel):
triggering_user_name: str | None = None
consumed_asset_events: list[AssetEventDagRunReference]
partition_key: str | None
+ note: str | None = None
+
+ @model_validator(mode="before")
+ @classmethod
+ def extract_dag_run_note(cls, data: Any) -> Any:
+ """Extract the `note` (`str | None` from
`association_proxy("dag_run_note", "content")`) relationship from `DagRun` to
prevent `DetachedInstanceError` when constructing `DagRunContext` or
`TIRunContext` models."""
+ from sqlalchemy import inspect as sa_inspect
+ from sqlalchemy.exc import NoInspectionAvailable
+ from sqlalchemy.orm.state import InstanceState
+
+ if isinstance(data, dict):
+ return data
+
+ # Check if this is a SQLAlchemy model by looking for the inspection
interface
+ try:
+ insp: InstanceState = sa_inspect(data)
+ except NoInspectionAvailable:
+ # Not a SQLAlchemy object, return as-is for Pydantic to handle
+ return data
+
+ # Check if dag_run_note is already loaded (avoid lazy load on detached
instance)
+ if "note" in insp.dict:
+ note_value: str | None = insp.dict["note"]
+ else:
+ note_value = None
+
+ # Convert to dict to avoid further lazy loading issues
+ values = {
+ field_name: getattr(data, field_name, None)
+ for field_name in cls.model_fields
+ if field_name != "note"
+ }
+ values["note"] = note_value
Review Comment:
Since we have introduced a `note` field to
`execution_api.datamodels.DagRun` in this PR, the `dag_run:
execution_api.datamodels.DagRun` attribute in `TIRunContext` and
`DagRunContext` now tries to access the `note` attribute on a `DagRun` SQLA
ORM instance, which leads to CI failures such as
`
Error extracting attribute: DetachedInstanceError: Parent instance DagRun at
0x112bdc6a0 is not bound to a Session; lazy load operation of attribute
dag_run_note cannot proceed (Background on this error at:
https://sqlalche.me/e/20/bhk3) [type=get_attribute_error, input_value=DagRun
test_dag @ 2026-0...ne. run_type: scheduled input_type=DagRun]
`.
To resolve these test errors, we could 1) let the Pydantic model respect
SQLA instance attributes, or 2) update all the upstream SQL statements to join
the `DagRunNote` table (for example,
`.options(selectinload(TI.dag_run).load_only(DagRun.note))`).
We have tried option 2) before, but it required changing SQL statements in
critical sections of the scheduler runner and other areas, which might not be
necessary. Therefore, I decided to go with option 1), based on the PoC at
https://github.com/jason810496/airflow/compare/fe19338…63d0095.
--
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]
