pierrejeambrun commented on code in PR #67242:
URL: https://github.com/apache/airflow/pull/67242#discussion_r3318602408
##########
airflow-core/src/airflow/api_fastapi/core_api/routes/ui/dags.py:
##########
@@ -308,3 +311,45 @@ def get_latest_run_info(dag_id: str, session: SessionDep)
-> DAGRunLightResponse
latest_run_info = session.execute(latest_run_info_select).one_or_none()
return DAGRunLightResponse(**latest_run_info._mapping) if latest_run_info
else None
+
+
+@dags_router.get(
+ "/run_state_counts",
+ dependencies=[
+ Depends(requires_access_dag(method="GET")),
+ Depends(requires_access_dag(method="GET",
access_entity=DagAccessEntity.RUN)),
+ ],
+ operation_id="get_dag_run_state_counts_ui",
+)
+def get_dag_run_state_counts(
+ session: SessionDep,
+ readable_dags_filter: ReadableDagsFilterDep,
+ dag_ids: Annotated[list[str], Query(min_length=1)],
+ run_after_gte: datetime | None = None,
+) -> DAGsRunStateCountsCollectionResponse:
+ """Return per-Dag DagRun state counts (zero-filled) for the Dag list
page."""
+ permitted_dag_ids = readable_dags_filter.value or set()
+ requested_dag_ids = [dag_id for dag_id in dict.fromkeys(dag_ids) if dag_id
in permitted_dag_ids]
+ counts_by_dag: dict[str, dict[DagRunState, int]] = {
+ dag_id: {state: 0 for state in DagRunState} for dag_id in
requested_dag_ids
+ }
+
+ if requested_dag_ids:
+ count_query = (
+ select(DagRun.dag_id, DagRun.state, func.count().label("cnt"))
+ .where(DagRun.dag_id.in_(requested_dag_ids))
+ .group_by(DagRun.dag_id, DagRun.state)
+ )
+ if run_after_gte is not None:
+ count_query = count_query.where(DagRun.run_after >= run_after_gte)
+ for row in session.execute(count_query):
+ if row.state is None:
+ continue
+ counts_by_dag[row.dag_id][DagRunState(row.state)] = row.cnt
+
Review Comment:
Both the home page and listing dag runs should work fast on a big install
(5M runs I think is big enough to highlight issues).
I think we focused our effort around big TI tables, and not on Runs table
that big, but that could be highlighting that the dag run part still needs more
improvement.
How do you generate that many runs?
> I wonder whether there's a better option than making most users see
limited information because of a design built around a tiny minority of heavy
users. What do you think?
UI should be scalable, for now there is no 'config' option, we just show
what is scalable and once that's done we can imagine adding a layer of
complexity to actually configure what information should be displayed, based on
the installation size. (maybe a high availability / scalability mode for the
api server, just throwing out ideas).
But for now we can't design pages that will crash when the airflow install
is big.
--
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]
