Skip to content

Limit deadline alerts endpoint to the latest serialized Dag row - #70804

Merged
vatsrahul1001 merged 5 commits into
apache:mainfrom
imrichardwu:fix/deadline-limit
Aug 3, 2026
Merged

Limit deadline alerts endpoint to the latest serialized Dag row#70804
vatsrahul1001 merged 5 commits into
apache:mainfrom
imrichardwu:fix/deadline-limit

Conversation

@imrichardwu

Copy link
Copy Markdown
Contributor

The GET /ui/dags/{dag_id}/deadlineAlerts endpoint queries serialized_dag for the latest serialized DAG by ordering rows in descending order, but the query never applies a LIMIT. Although session.scalar() returns a single object, it does not limit the SQL query itself. The database and ORM still materialize the full result set before discarding every row except the first.

Since every serialized_dag row contains the full serialized DAG blob, this becomes a serious scalability issue for DAGs with long version histories. A DAG with thousands of historical versions can force the API server to load gigabytes of serialized data into memory just to return the newest entry. In one reported case, a DAG with roughly 6,725 versions caused the API process to consume about 4 GiB of memory and exceed its memory limit.

Adding .limit(1) ensures the database returns only the newest row. This reduces memory usage from scaling with a DAG's version history to a constant amount, while also allowing the query planner to stop after the first matching index entry instead of processing every matching row.

This also makes the implementation consistent with the existing _get_latest_serdag helper in grid.py, which already performs the same lookup with LIMIT 1.

closes: #70779


Was generative AI tooling used to co-author this PR?
  • Yes (please specify the tool below)

  • Read the Pull Request Guidelines for more information. Note: commit author/co-author name and email in commits become permanently public when merged.
  • For fundamental code changes, an Airflow Improvement Proposal (AIP) is needed.
  • When adding dependency, check compliance with the ASF 3rd Party License Policy.
  • For significant user-facing changes create newsfragment: {pr_number}.significant.rst, in airflow-core/newsfragments. You can add this file in a follow-up commit after the PR is created so you know the PR number.

ordering all rows for a dag_id descending and reading the first one.
Without a LIMIT, the database materializes every serialized version of
the Dag when only the newest is needed. Cap the lookup at a single row.
Comment thread airflow-core/src/airflow/api_fastapi/core_api/routes/ui/deadlines.py Outdated
@eladkal eladkal added this to the Airflow 3.3.1 milestone Jul 31, 2026
@eladkal eladkal added type:bug-fix Changelog: Bug Fixes backport-to-v3-3-test Backport to v3-3-test labels Jul 31, 2026
@eladkal
eladkal requested a review from ferruzzi July 31, 2026 11:26

@ferruzzi ferruzzi left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Left a non-blocking suggestion in Elad's comment thread, otherwise LGTM

A backfill run executes the Dag version that was current when it was
queued, whose deadline alerts may differ from the ones deployed today.
The endpoint could only ever resolve the newest serialized Dag, so the
UI reported present-day alerts for historical runs and misrepresented
what those runs were actually held to.

Resolving the version through DagVersion.version_number also removes a
reliance on the serialized_dag primary key sorting chronologically,
which only held because uuid7 happens to be time-ordered and would give
the wrong row if an older version were serialized after a newer one.
Only the primary key is read, so the serialized Dag blob no longer
crosses the wire for a lookup that discards it.
@imrichardwu

Copy link
Copy Markdown
Contributor Author

I updated it based on your suggestion @ferruzzi

@vatsrahul1001
vatsrahul1001 merged commit 1b3121d into apache:main Aug 3, 2026
99 checks passed
@github-actions

github-actions Bot commented Aug 3, 2026

Copy link
Copy Markdown
Contributor

Backport successfully created: v3-3-test

Note: As of Merging PRs targeted for Airflow 3.X
the committer who merges the PR is responsible for backporting the PRs that are bug fixes (generally speaking) to the maintenance branches.

In matter of doubt please ask in #release-management Slack channel.

Status Branch Result
v3-3-test PR Link

bbovenzi pushed a commit that referenced this pull request Aug 3, 2026
…ag row (#70804) (#70965)

* The deadline alerts endpoint fetches the latest serialized Dag by
ordering all rows for a dag_id descending and reading the first one.
Without a LIMIT, the database materializes every serialized version of
the Dag when only the newest is needed. Cap the lookup at a single row.

* Sort imports in deadline alerts UI route tests

* UI: Allow reading deadline alerts of a specific Dag version

A backfill run executes the Dag version that was current when it was
queued, whose deadline alerts may differ from the ones deployed today.
The endpoint could only ever resolve the newest serialized Dag, so the
UI reported present-day alerts for historical runs and misrepresented
what those runs were actually held to.

Resolving the version through DagVersion.version_number also removes a
reliance on the serialized_dag primary key sorting chronologically,
which only held because uuid7 happens to be time-ordered and would give
the wrong row if an older version were serialized after a newer one.
Only the primary key is read, so the serialized Dag blob no longer
crosses the wire for a lookup that discards it.
(cherry picked from commit 1b3121d)

Co-authored-by: Richard <richard9@ualberta.ca>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

area:API Airflow's REST/HTTP API backport-to-v3-3-test Backport to v3-3-test type:bug-fix Changelog: Bug Fixes

Projects

None yet

Development

Successfully merging this pull request may close these issues.

UI deadlineAlerts endpoint loads all historical serialized DAG rows

4 participants