kezhuw commented on a change in pull request #15605:
URL: https://github.com/apache/flink/pull/15605#discussion_r613867235
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/operators/coordination/SubtaskGatewayImpl.java
##########
@@ -64,6 +69,17 @@
subtaskAccess.createEventSendAction(serializedEvent);
final CompletableFuture<Acknowledge> result = new
CompletableFuture<>();
+ result.whenCompleteAsync(
Review comment:
To be precise, the "correctness" above should be accuracy(or similar
words) of pending future tracking, it does not contribute to exactly-once
guarantee of coordinator holder.
I constructed a test case with following changes:
* Change `FutureUtils.forward(sendResult, result)` in
`OperatorEventValve.callSendAction` to a delayed forward.
* Fail subtask but never cancel event sending with outdated gateway.
Then I see more checkpoint abortion than usual.
--
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]
