rkhachatryan commented on a change in pull request #14024:
URL: https://github.com/apache/flink/pull/14024#discussion_r521220231
##########
File path:
flink-tests/src/test/java/org/apache/flink/test/checkpointing/UnalignedCheckpointITCase.java
##########
@@ -531,8 +534,10 @@ public void invoke(Long value, Context context) throws
Exception {
state.numOutput++;
if (state.completedCheckpoints < minCheckpoints) {
- // induce heavy backpressure until enough
checkpoints have been written
- Thread.sleep(0, 100_000);
+ // induce backpressure until enough checkpoints
have been written
+ if (random.nextInt(1000) == 42) {
Review comment:
> Why do we need to reduce the backpressure here?
This is a commit from #13827. As far as I understand, the motivation was:
1. to have higher backpressure (.1ms vs 1ms) for **some** parts of the stream
2. to have unbalanced streams
3. reduce running time (was ~1m, now ~10s)
On average, the change decreases sleep time only by 60% (1-42/1000*10).
> I'm worried that any solution on top of it will only work with a certain
minimal flow.
With 1K elements, the sink will sleep 23 times for 1ms, which I think is
enough and 1K is a reasonable number of elements for this test.
----------------------------------------------------------------
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]
