Github user jinxing64 commented on the issue:
https://github.com/apache/spark/pull/18388
@tgravescs
Thanks a lot for helping this pr !
I changed this pr, in current change:
Shuffle server will track the number of chunks being transferred.
Connection will be closed when the number is above water mark.
I think it makes sense -- shuffle server should close the connection rather
than OOM when memory pressure is high.
For the flow control part, it is not included in current change. Because I
don't have a proper way to balance between buffering requests and caching
`ChannelOutboundBuffer$Entry`.
>You mentioned a problem with just doing a close, what was the problem?
I just think user should know there's another condition that can trigger
closing connection. Also reconnection can have some cost.
---
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
