wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1024817221
I still believe this is a working fix. And no more complex than yours. They
are on the same level.
If the decision is made, then all I can say is that it takes a lot efforts
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1024815295
If the `Complete` indicates the input future completion. (Thats a little
different in my graph. In the graph it refers to the line the calls the
`complete` on the result
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1024448903
I can make test for the scenario you described which my fix will fail. Just
manually create a gap, resolve the reference first, then a long sleep, then try
to complete. During
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-102960
In your version, under the same circumtance, `anyAvailable.get()` may get
the obsolete object either. `AtomicReference` can not guranteen the
`happen-before` relationship you
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1024372868
> As far as I can tell, the case in the 3rd chart is not working in your
version. If `Complete` action happens at the very very end (around `to here`
arrow), there is no
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1024349502
BellSoft Liberica jdk8u292.
https://user-images.githubusercontent.com/2689362/151578190-851c24c5-f7ba-4f4d-b1b7-03ea96e45fd6.png;>
--
This is an automated message from
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1024191941
I made a graph to explain how this fix works. Even with the race condition
between future's `notifyCompletion` and the `checkReusableAndReset`.
![flink-issue-timeline
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-102405
>
> @wpc009 . Could you check my PR if it's solving the memory leak? My
proposal is quite a bit simpler.
>
> Regarding the tests, unfortunately I don't
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1024041597
> @wpc009 . Could you check my PR if it's solving the memory leak? My
proposal is quite a bit simpler.
>
> Regarding the tests, unfortunately I don't see a way for
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1023958486
Here is some benchmark results (From my laptop)
| Benchmark
| Mode | Cnt | Score| Error | Units
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1023885086
I have a fix.
Tested on the `flink-benchmark`. I'm preparing some end-to-end test. I can
push it at the end of today.
--
This is an automated message from the Apache Git
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1022813985
It seems like newer JDK release have already fixed the memory issue of
`CompletableFuture.anyOf`.
Saddly, this issue is still exists in jdk8. It's time to abandon this old
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1021816172
Maybe I should push to `release-1.14` branch?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1021815826
The master branch can not pass the CI pipeline either.
Check this out.
https://dev.azure.com/wpc009/Flink/_build/results?buildId=5=results
--
This is an automated
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1021431119
> I'm afraid that those "time outs" might be because of some bug in this
code causing a deadlock. I've also tried [to run a benchmark
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1020772870
> ## CI report:
> *
[3ade971](https://github.com/apache/flink/commit/3ade9719b565beeacf7761ab71d5abb7ba62e62a)
Azure:
wpc009 commented on pull request #18475:
URL: https://github.com/apache/flink/pull/18475#issuecomment-1020772870
> ## CI report:
> *
[3ade971](https://github.com/apache/flink/commit/3ade9719b565beeacf7761ab71d5abb7ba62e62a)
Azure:
17 matches
Mail list logo