The Windows CI fails due to timeout.

[HideakiImamura]

Expected behavior

I have been observing an issue for the past few days where the Windows CI randomly fails due to timeout, regardless of whether it's a PR or on schedule. It looks something like the following CI:　https://github.com/optuna/optuna/actions/runs/6777581250/job/18421473256#step:8:15

All of the CIs appear to fail on the first test case of the visualization features using matplotlib. I have no windows machine to reproduce the error on my local environment.

Environment

• Optuna version: the current master 523dcc2
• Python version: 3.9
• OS: Windows Server 2022
• (Optional) Other libraries and their versions:

Error messages, stack traces, or logs

Here is the CI's output.

Run pytest -m "not integration"
============================= test session starts =============================
platform win32 -- Python 3.9.13, pytest-7.4.3, pluggy-1.3.0
rootdir: D:\a\optuna\optuna
configfile: pyproject.toml
plugins: jaxtyping-0.2.23, typeguard-2.13.3
collected 4850 items / 839 deselected / 4011 selected

tests\test_callbacks.py .                                                [  0%]
tests\test_cli.py ...................................................... [  1%]
...
tests\trial_tests\test_trial.py ........................................ [ 80%]
........................................................................ [ 81%]
..............................                                           [ 82%]
tests\trial_tests\test_trials.py ....................................... [ 83%]
...                                                                      [ 83%]
tests\visualization_tests\test_contour.py ..
Error: The operation was canceled.

Steps to reproduce

Trigger the windows CI by workflow dispatcher on your fork.

Additional context (optional)

No response

[not522]

CI seems to be stuck on the third test in test_contour.py. This is the first test in test_plot_contour, which outputs an image using plotly. I am not sure, but perhaps this issue plotly/Kaleido#110 is related.

[HideakiImamura]

Downgrading kaleido to 0.1.0 will resolve issue according to the comments like plotly/Kaleido#110 (comment). I will hotfix it.

[not522]

According to the issue, 0.1.0post1 may be better.

[not522]

Unfortunately, kaleido 0.1.0post1 did not resolve this issue. https://github.com/optuna/optuna/actions/runs/6818449794

[not522]

In my understanding, there were two problems related to this issue, test_contour.py and test_journal.py. The first one was resolved by Kaleido 0.1.0post1. The second one has not been reproduced in these two months. I think that it is OK to close this issue. If this issue will be reproduced or you have some concerns, please feel free to reopen it.

[not522]

I reopen this issue since the problem with test_journal.py recurred. https://github.com/optuna/optuna/actions/runs/8530596405