The GitHub Actions job "Tests (AMD)" on airflow.git/fix/57512-composer-dag-run-sensor-empty-window has failed. Run started by GitHub user Vamsi-klu (triggered by Vamsi-klu).
Head commit for run: 66518cac4b664d1c30702cfb981c6cdac7da829e / deepinsight coder <[email protected]> Pin CloudComposer system test to triggered run id Reorder the example Dag so each CloudComposerTriggerDAGRunOperator runs before its sibling CloudComposerDAGRunSensor, and have the sensors pull the freshly minted run id from XCom via the new templated composer_dag_run_id field. The previous "execution_range=[now -1d, now]" form was evaluated at Dag-parse time -- before the Composer environment existed -- so the empty-window fix from PR #67052 would have caused the sensor to wait forever, the same way it did when PR #61046 was reverted. Adds a defensive timeout on the Dag-run sensors so any future regression back into the windowed code path fails CI fast instead of hanging the worker. Leaves the external-task sensors on the legacy windowed path deliberately; the analogous bug in _check_task_instances_states is tracked at https://github.com/apache/airflow/issues/67051 and the inline note flags that the example must be restructured at the same time when that issue lands. Adds unit coverage for: the new template_fields entry, Jinja rendering of an xcom_pull template, run-id-path success/wait/precedence in poke(), parse-time source guard against re-introducing datetime.now(), and the matching trigger paths (run-id branch precedence + polling). Ships a local simulation harness (dev/simulate_composer_system_test.py) that mocks all GCP surfaces and drives the example via dag.test() end-to-end so contributors without GCP credentials can reproduce the proof. Report URL: https://github.com/apache/airflow/actions/runs/26236628942 With regards, GitHub Actions via GitBox --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
