hussein-awala commented on issue #75865: URL: https://github.com/apache/airflow/issues/75865#issuecomment-6521424245
Tested my changes against **Airflow 3.4.0rc1 / Task SDK 1.4.0rc1**, using Breeze with Colima, PostgreSQL 14, and LocalExecutor. I prepared data and exercised real Dag scenarios: **5 Dags, 6 successful runs, and 27 successful task instances**. - #64610: Exact and regex partition-key filtering worked through both the public REST API (`GET /api/v2/assets/events`) and Task SDK `inlet_events`, including combined conditions that correctly returned no matches. - #64611: Extra-field AND filtering worked through both the public REST API and Task SDK `inlet_events`. Inlet checks also covered alias events, missing matches, and literal dotted keys. - #74498: The dry-run preserved data. Cleanup removed the selected asset’s old event while retaining another asset’s event older than the cutoff and the selected asset’s newer event. - #74592: A waiting task adopted the new Dag version before execution, while the completed task retained its original version. Their outputs confirmed `v1` and `v2`, respectively. - #72848: With two healthy triggerers, callback ownership stayed stable across 20 seconds of observation. There was one callback start, one completion, and no cancellation. All these scenarios passed. Version handling used a local unversioned Dag bundle; triggerer failover was not exercised. During initial preparation, concurrent first creation of the same alias target hit `idx_asset_name_uri_unique`. Declaring the target asset before execution allowed the scenarios to complete; this observation may warrant separate investigation. -- 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]
