[CI][C++] arrow-compute ExecPlanExecution.StressSourceGroupedSumStop timeout

[mapleFU]

Describe the bug, including details regarding any error messages, version, and platform.

Test is: https://github.com/apache/arrow/actions/runs/3862492798/jobs/6584098719

[ RUN      ] ExecPlanExecution.StressSourceGroupedSumStop
D:/a/arrow/arrow/cpp/src/arrow/compute/exec/plan_test.cc:849: Failure
Value of: _fut.Wait(::arrow::kDefaultAssertFinishesWaitSeconds)
  Actual: false
Expected: true
Google Test trace:
D:/a/arrow/arrow/cpp/src/arrow/compute/exec/plan_test.cc:824: single threaded
D:/a/arrow/arrow/cpp/src/arrow/compute/exec/plan_test.cc:821: slowed
D:/a/arrow/arrow/cpp/src/arrow/compute/exec/exec_plan.cc:64: Plan was destroyed before finishing


99% tests passed, 1 tests failed out of 69

which cause this patch failed #15241

Component(s)

C++, Continuous Integration

[rok]

Looks like the test takes too long to run?

[mapleFU]

Looks like the test takes too long to run?

Maybe just running to much time / a deadlock or other bug...I don't know :(

[westonpace]

kDefaultAssertFinishesWaitSeconds is generally long enough that even CI machines will finish by then. So I think deadlock seems likely :(

[westonpace]

I've reproduced this. It is a deadlock. It appears that, in some cancel situations, GroupByNode is not properly marking it's finished future as complete. I can probably spot fix this if we need to but it is fixed more categorically by #15253 (I'm attempting to reproduce on that branch and so far have not been able to so that is a good sign). If it appears that #15253 is not going to merge before the release then I will fix this individually (feel free to ping me if it seems like I am forgetting).

[mapleFU]

Please go ahead, I just blocked by CI

By the way, I meet another timeout, but I don't understand why: https://github.com/apache/arrow/actions/runs/3886295191/jobs/6631229773

[westonpace]

I don't think #15253 will merge in time. I am going to try and put in a more targeted fix today.

[mapleFU]

Hi westonpace, did you solve this issue?

[westonpace]

Hi westonpace, did you solve this issue?

Thanks for the reminder, I had forgotten.

[westonpace]

@raulcd this should be included in the release. I've marked it with the 11.0.0 milestone. Is there anything more I need to do?

[raulcd]

no, that's fine, Thanks! The archery release cherry-pick script will pick it up. I am adding the Priority: blocker label, just as a reminder for me to double check once I run the cherry-pick command but adding the 11.0.0 milestone should be enough.