pnowojski commented on a change in pull request #7911: [FLINK-11082][network]
Fix the logic of getting backlog in sub partition
URL: https://github.com/apache/flink/pull/7911#discussion_r265985944
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/io/network/partition/SpilledSubpartitionView.java
##########
@@ -147,7 +147,8 @@ public BufferAndBacklog getNextBuffer() throws
IOException, InterruptedException
return null;
}
- int newBacklog =
parent.decreaseBuffersInBacklog(current.isBuffer());
+ parent.decreaseBuffersInBacklog(current.isBuffer());
+ int newBacklog = parent.getBuffersInBacklog();
Review comment:
I think that option 1. even if it's correct right now, would be more error
prone for the future. Forcing every contributor/committer in this part to
rethink this contract/dependencies every time is not a good thing to do.
For example, as you if the backlog increases between those two calls there
is no problem, however it takes some time to realize about this fact. Also
currently I'm not sure if the backlog can be somehow decreased by a 3rd party
(another thread) betwen those two calls and if that happens, if that's ok?
Another question, are all of the arithmetic operations on the backlog atomic?
Long story short, if we can avoid think about the above issues, so much the
better - I would pick number 2.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
