akalash commented on a change in pull request #11877:
URL: https://github.com/apache/flink/pull/11877#discussion_r662841374
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/io/network/partition/PipelinedSubpartition.java
##########
@@ -288,9 +288,7 @@ BufferAndBacklog pollBuffer() {
if (buffers.isEmpty()) {
flushRequested = false;
- }
-
- while (!buffers.isEmpty()) {
+ } else {
Review comment:
Yes, the batch execution mode still requires the manual calling of
finish which brokes my idea. The same about PipelinedApproximateSubpartition.
As I understand, it is impossible to know that a partial record will be sent or
not before the request the credit.
So I still believe that it would be better to avoid sending the empty buffer
if it is possible. But perhaps you right and it will be not so easy to do so in
the current implementation, especially because of the scenarios which you
described.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
