Github user kayousterhout commented on the issue:
https://github.com/apache/spark/pull/16620
BTW Mark one slightly different version of your suggestion I'd considered
is:
(1) move stage.pendingPartitions -= task.partitionId so that it's
duplicated in each of the two case statements below
(2) for the ResultTask case, removing the partition can happen right at the
beginning
(3) for the ShuffleMapTask case, removing the partition can happen in the
else statement on line 1196, where addOutputLoc is called.
One benefit of that approach is that it makes it a little more obvious
which state is related: that the pendingPartitions should mirror (in reverse)
the output locations for the state. It also consolidates the logic for
handling previously failed executors into the one location.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
