houqp commented on a change in pull request #6342: [AIRFLOW-5662] fix incorrect
naming and batch db call for scheduler metrics
URL: https://github.com/apache/airflow/pull/6342#discussion_r405032075
##########
File path: airflow/models/pool.py
##########
@@ -62,6 +71,45 @@ def get_default_pool(session=None):
"""
return Pool.get_pool(Pool.DEFAULT_POOL_NAME, session=session)
+ @staticmethod
+ @provide_session
+ def slots_stats(session=None) -> Dict[str, PoolStats]:
+ from airflow.models.taskinstance import TaskInstance # Avoid circular
import
+
+ pools: Dict[str, PoolStats] = {}
+
+ pool_rows: Iterable[Tuple[str, int]] = session.query(Pool.pool,
Pool.slots).all()
+ for (pool_name, total_slots) in pool_rows:
+ pools[pool_name] = PoolStats({
+ "total": total_slots,
+ State.RUNNING: 0,
+ State.QUEUED: 0,
+ })
+
+ state_count = (
+ session.query(TaskInstance.pool, TaskInstance.state, func.count())
+ .filter(TaskInstance.state.in_(list(EXECUTION_STATES)))
+ .group_by(TaskInstance.pool, TaskInstance.state)
+ ).all()
Review comment:
i have removed the if state check in that for loop, it's not needed anymore
with explicit state filter in the query.
with regards to your proposed change: if we remove group by state from the
query, how do we get queued slot count for a particular pool?
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
