Github user kayousterhout commented on the pull request:
https://github.com/apache/spark/pull/6423#issuecomment-111307212
My only remaining concern is with the unit test: we used to check that the
buffer got released when the iterator was exhausted. Now, we test that the
buffer is released when close() gets called on the wrapped stream, but we don't
verify that close() ever gets called. It would be great to do the latter, just
because not releasing the input stream can lead to nasty OOM bugs (particularly
because this data is off-heap in a managed buffer). You'd need to add a test
for HashShuffleReader for that (and exhaust that iterator, similar to what the
old test for ShuffleBlockFetcherIterator did); is that prohibitively difficult?
I also pinged @rxin who said he'd take a look at this tonight -- need
sign-off from him or another shuffle maintainer since this is a non-tiny change
to the shuffle code.
---
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]
