Github user tgravescs commented on the issue:
https://github.com/apache/spark/pull/21558
yeah that is the code that I was still looking at to verify if it can
actually happen. on a fetch failure the dag scheduler removes the stage from
the ouput committer, but when the new stage attempt starts it just puts the
stage back with the stage id (not attempt id).
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
