You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently we're observing ongoing flakiness of XGBoost sample. A brief investigation shows that it's highly likely caused by the fact that we currently do not have a way to correctly specify the new GCP dataproc_delete_cluster as exit handler component, then argo cannot determine the execution order properly, resulting in race condition.
What do we need:
This should be fixed by modifying the corrensponding component yaml spec. For example, the generated ContainerOp should have is_exit_handler = True
The text was updated successfully, but these errors were encountered:
Currently we're observing ongoing flakiness of XGBoost sample. A brief investigation shows that it's highly likely caused by the fact that we currently do not have a way to correctly specify the new GCP dataproc_delete_cluster as exit handler component, then argo cannot determine the execution order properly, resulting in race condition.
What do we need:
This should be fixed by modifying the corrensponding component yaml spec. For example, the generated ContainerOp should have is_exit_handler = True
The text was updated successfully, but these errors were encountered: