Repository navigation
Conversation
In Airflow 3, SerializedDagModel stores serialized definitions per DagVersion, accumulating large JSON and binary blobs as DAG versions evolve. Previously, serialized_dag was not included in db clean table configurations, preventing operators from purging unreferenced historical serialized DAG records and cascading deadline alerts.
zozo123
left a comment
zozo123
left a comment
There was a problem hiding this comment.
LGTM. Deferring to maintainers for final 👍
| recency_column_name="created_at", | ||
| extra_columns=["id"], | ||
| dependent_tables=["task_instance", "dag_run"], | ||
| dependent_tables=["task_instance", "dag_run", "serialized_dag"], |
May I ask what Airflow version do you use? (it might been solved by #68339, release in v3.3.0) |
|
Closing because #68339 already implements the correct cleanup path through Drafted-by: GitHub Copilot (GPT-5.6) (no human review before posting) |
Include
serialized_daginairflow db cleantable configurations.In Airflow 3,
SerializedDagModelstores serialized definitions perDagVersion, accumulating JSON and binary data over time. Previously,serialized_dagwas omitted fromdb clean, preventing operators from purging unreferenced historical serialized DAG records.This change:
serialized_dagtoconfig_listindb_cleanup.pywith recency tracking oncreated_atand version reference checks.keep_last=Truegrouped bydag_idto preserve the latest serialized DAG definition.serialized_dagtodag_versiondependent tables so serialized records and cascadingdeadline_alertrows are cleaned and archived before version records.Was generative AI tooling used to co-author this PR?
Generated-by: Gemini 3.8 following the guidelines