uranusjr commented on code in PR #71072:
URL: https://github.com/apache/airflow/pull/71072#discussion_r3793714706
##########
airflow-core/src/airflow/jobs/scheduler_job_runner.py:
##########
@@ -2269,6 +2279,83 @@ def _create_dagruns_for_partitioned_asset_dags(self,
session: Session) -> set[st
if not pending_apdrs:
return set()
+ if len(pending_apdrs) >= self._max_partition_dag_runs_per_loop:
+ backlog_total = (
+ session.scalar(
+ select(func.count())
+ .select_from(AssetPartitionDagRun)
+ .join(DagModel, DagModel.dag_id ==
AssetPartitionDagRun.target_dag_id)
+ .where(
+ AssetPartitionDagRun.created_dag_run_id.is_(None),
+ DagModel.is_stale.is_(False),
+ )
+ )
+ or 0
+ )
+ # Distinct dag_ids across the *whole* backlog, not just this
tick's oldest-cap
+ # slice (`pending_apdrs`) — a Dag whose partitions haven't reached
the front of
+ # the FIFO queue yet would otherwise be missing from the log/audit
row until its
+ # turn comes up. Same WHERE clause as the count query above so the
two never drift.
+ backlog_dag_ids = set(
+ session.scalars(
+ select(AssetPartitionDagRun.target_dag_id)
+ .join(DagModel, DagModel.dag_id ==
AssetPartitionDagRun.target_dag_id)
+ .where(
+ AssetPartitionDagRun.created_dag_run_id.is_(None),
+ DagModel.is_stale.is_(False),
+ )
+ .distinct()
+ )
+ )
+ else:
+ backlog_total = len(pending_apdrs)
+ self._partition_cap_backlog_reported = False
+
+ if backlog_total > self._max_partition_dag_runs_per_loop:
+ self.log.warning(
+ (
+ "Reached the per-tick cap on pending partitioned Dag runs;
the remaining backlog "
+ "will be evaluated over subsequent scheduler ticks"
+ ),
+ cap=self._max_partition_dag_runs_per_loop,
+ backlog_total=backlog_total,
+ dag_ids=backlog_dag_ids,
+ )
+ # Edge-trigger the audit row: a persistent backlog re-hits this
branch every tick,
+ # and writing a `Log` row that often would flood the audit table.
Write it once per
+ # backlog episode; `_partition_cap_backlog_reported` is cleared
below once the
+ # backlog drains.
+ if not self._partition_cap_backlog_reported:
+ # A separate, independently-committed session is required
here: the caller
+ # (`_create_dagruns_for_dags`) runs under
`@retry_db_transaction`, which rolls
+ # back *session* on a `DBAPIError` — sharing that transaction
would silently
+ # discard this audit row along with the rest of the tick's
work.
+ # Invariant: this must run before *session* makes any writes
this tick, or the
+ # new connection's commit can lock-contend with it on SQLite.
+ try:
+ with create_session(scoped=False) as audit_session:
+ audit_session.add(
+ Log(
+ event="partition Dag run cap reached",
+ extra=(
+ f"The scheduler evaluated
{len(pending_apdrs)} pending partitioned Dag "
+ f"runs this tick, reaching the internal
per-tick cap of "
+ f"{self._max_partition_dag_runs_per_loop}.
The remaining backlog will "
+ f"be evaluated over subsequent ticks. A
total of {backlog_total} "
+ f"partitioned Dag runs are currently
eligible and pending across "
+ f"ticks. Affected dag_ids: {',
'.join(backlog_dag_ids)}."
Review Comment:
Also we should probably sort the ids here to make the message stable.
--
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]
