Update distributed CI test dependencies

[jrbourbeau]

This PR adds pytest-asyncio and pytest-rerunfailures as dependencies to distributed downstream CI build.

pytest-asyncio is needed to run some distributed tests, otherwise they will be skipped. For example, looking at the CI builds in #432, we see this warning about being skipped for lots of tests:

...
distributed/tests/test_utils.py: 1 warning
distributed/tests/test_utils_test.py: 4 warnings
distributed/tests/test_worker.py: 7 warnings
  /opt/hostedtoolcache/Python/3.7.11/x64/lib/python3.7/site-packages/_pytest/python.py:172: PytestUnhandledCoroutineWarning: async def functions are not natively supported and have been skipped.
  You need to install a suitable plugin for your async framework, for example:
    - anyio
    - pytest-asyncio
    - pytest-tornasync
    - pytest-trio
    - pytest-twisted
    warnings.warn(PytestUnhandledCoroutineWarning(msg.format(nodeid)))

pytest-rerunfailures was recently added as a test dependency for distributed to automatically re-run some known flaky tests.

cc @jakirkham

[codecov]

Codecov Report

Merging #445 (ac52c58) into master (fca7b22) will increase coverage by 0.16%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #445      +/-   ##
==========================================
+ Coverage   92.36%   92.52%   +0.16%     
==========================================
  Files           4        4              
  Lines         720      709      -11     
  Branches      150      155       +5     
==========================================
- Hits          665      656       -9     
+ Misses         34       32       -2     
  Partials       21       21              

Impacted Files	Coverage Δ
cloudpickle/__init__.py	100.00% <0.00%> (ø)
cloudpickle/cloudpickle.py	88.53% <0.00%> (+0.22%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fca7b22...ac52c58. Read the comment docs.

[jakirkham]

Thanks James! 😀

Should we start running these skipped tests again?

Also added a label so the next commit pushed will run the Distributed test suite.

[jrbourbeau]

Also added a label so the next commit pushed will run the Distributed test suite.

👍

Should we start running these skipped tests again?

Hrm good question. It should be safe to run those tests again since they've both been marked as flaky in distributed and should be automatically rerun. I'll uncomment the current skips to confirm they pass here. I'm also going to add the --runslow flag used in distributed to unskip a bunch of other tests. Let's see how much longer that makes the distributed CI build take to run

[jrbourbeau]

I decided to punt on running all slow tests in distributed for now since there are still a few flaky tests causing issues here. However, we can start running test_dont_steal_unknown_function and test_target_duration as they've been taken care of in distributed

[jakirkham]

Thanks for the update James! 😄 Yeah that makes sense

It looks like CI is encountering a couple of test failures. In particular test_profile_server and test_log_exception_on_failed_task. Do we know what is going on with those?

[jrbourbeau]

Marking test_profile_server as flaky over in dask/distributed#5375. test_log_exception_on_failed_task is marked with avoid_cidownstream in distributed, so I pushed a commit here to avoid all tests marked with avoid_ci (TIL this is what we do in distributed). Let's see if that gives us a green check mark 🤞

[jrbourbeau]

Marking test_worker_reconnects_mid_compute_multiple_states_on_scheduler as flaky in dask/distributed#5378

[jakirkham]

Just realized we hadn't re-run after merging that. Trying to rerun CI

[jrbourbeau]

Woo, all green ✅

Unfortunately distributed has several flaky tests right now. I'm going to push a couple of empty commits to see if repeated CI runs trigger another failure or not. FWIW I'm not sure how often downstream CI is run here, but if an unexpected failure pops up feel free to ping me. I'm happy to help out

[jakirkham]

Looks like the last commit also worked. Was there anything else you wanted to do here James? Or are we good to merge?

I think it is run mainly before releases, but may be run on other changes if there’s concerns that the changes will affect downstream projects

[jakirkham]

Hmm...actually now seeing test_scheduler_address_config fail

[jrbourbeau]

That's unfortunate : / I think we'll continue to have some flaky tests here in the distributed build until we're able to mark as flaky / xfail / fix tests upstream in distributed. There's an ongoing effort to run the distributed test suite many times to get a better since for what tests are flaky (xref dask/distributed#5378 (comment)). Based on those results we can then go through an start targeting specific tests. My sense is the changes here are a net benefit and this PR is probably good to merge.

[jakirkham]

@pierreglaser @ogrisel , do you have thoughts here?

[pierreglaser]

LGTM, thanks a lot @jrbourbeau @jakirkham. Not much thoughts on this, apart from me appreciating your work on making the distributed build more robust to random failures.

Will wait a few days (if @ogrisel wants to comment), and then feel free to merge @jakirkham

[ogrisel]

LGTM, thanks a lot @jrbourbeau @jakirkham. Not much thoughts on this, apart from me appreciating your work on making the distributed build more robust to random failures.

Same feeling. Let's merge.

[ogrisel]

Thanks again @jrbourbeau!

[jakirkham]

Thanks all! 😄

[jrbourbeau]

Thanks all 🚀