Github user kayousterhout commented on the issue:
https://github.com/apache/spark/pull/17088
Can you update the JIRA and PR description to say "un-register the output
locations" (or similar) instead of "remove the files"? The current description
is misleading since nothing is actually getting removed from the machine --
instead you're just updating the scheduler's state.
Also, does the issue here only arise when the shuffle service is enabled?
It seems like when the shuffle service is not being used,
DAGScheduler.handleExecutorLost should be called for each lost executors (i.e.,
for all of the executors on a host, when a whole host is lost), which will
correctly un-register all of the output on that executor.
---
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]
