Github user mccheah commented on the pull request:
https://github.com/apache/spark/pull/8007#issuecomment-134468874
Hm, initial testing seems ok. I did run into a case where an asking was
lost. Not immediately clear as to why, but it turned out that the executor had
been removed anyways and the subsequent attempt to remove the now-non-existant
executor was harmless.
You can hit a race condition where the AM reporter thread processes the
disconnected executor before the onDisconnected-get-executor-loss-reason
request makes it to the AM, but that's probably okay since the AM would have
also sent a removeExecutor request to the driver with the right loss reason
anyways...
It would be nice to have an automated way to test this, but I don't
currently see any test suite for YarnSchedulerBackend and it looks fairly
difficult to conjure one.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
