Failure of one task will stop other tasks from being scheduled?

["EKC (Erik Cederstrand)" <EK...@novozymes.com>]

Hi folks,


We're hitting an issue where some of our tasks in a DAG run are never scheduled, even though all upstream tasks have succeeded. When the DAG run completes (with status=failed), these tasks just show up in the graph with state=null. This seems to happen only when at least one task in the entire DAG run has failed.


My suspicion is that when one task in the DAG run fails, the scheduler stops scheduling remaining tasks in this DAG run, even though the failing task is not a dependency for the remaining tasks. If so, is this intended behavior?


Our DAG consists of a range of data synchronization tasks for various unrelated systems, so we never retry the DAG, but we'd like to try to run *all* possible tasks even if some unrelated task fails.


Thanks,

Erik