The GitHub Actions job "Tests (AMD)" on 
airflow.git/feature/k8s-executor-pre-execution-requeue has failed.
Run started by GitHub user seanmuth (triggered by seanmuth).

Head commit for run:
87169b78a536e3a80465d2a3a22b87292c18a263 / seanmuth <[email protected]>
Requeue KubernetesExecutor tasks whose pod failed before execution started

When a worker pod is destroyed before the task process starts (node drain, 
autoscaler scale-down, node boot race, transient image pull failure), the task 
instance is still queued and no task code has run. Reporting this to the 
scheduler as a normal failure consumes a user-configured retry and raises a 
misleading failure alert for work that never executed. The executor already has 
the signal to tell this apart from an execution failure, so it now 
transparently requeues the pod without consuming a task retry, bounded by 
pod_launch_failure_retries and excluding container reasons in 
pod_launch_failure_excluded_container_reasons (default Error).

Report URL: https://github.com/apache/airflow/actions/runs/28263112577

With regards,
GitHub Actions via GitBox


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to