[
https://issues.apache.org/jira/browse/FLINK-34202?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17818269#comment-17818269
]
Xingbo Huang commented on FLINK-34202:
--------------------------------------
I reviewed all the stages where timeouts occurred and found that these stages
all ran on AlibabaCI001. Simultaneously, the runtime of all other successful
stages is consistently around 2 hours and 40 minutes. In the logs, I didn't
notice any tests being stuck or having an overly long runtime, so I think the
timeout is largely due to AlibabaCI001's performance not being sufficient to
complete 4 Python version tests within 4 hours. I submitted a PR to have the
nightly CI randomly select a Python version for testing rather than running all
4 Python versions. By only running one Python version test, even if the
machine's performance is poor, it should not exceed 2 hours (the lab triggered
by the PR only runs the latest Python version, which takes about 40 minutes).
> python tests take suspiciously long in some of the cases
> --------------------------------------------------------
>
> Key: FLINK-34202
> URL: https://issues.apache.org/jira/browse/FLINK-34202
> Project: Flink
> Issue Type: Bug
> Components: API / Python
> Affects Versions: 1.17.2, 1.19.0, 1.18.1
> Reporter: Matthias Pohl
> Assignee: Xingbo Huang
> Priority: Critical
> Labels: pull-request-available, test-stability
>
> [This release-1.18
> build|https://dev.azure.com/apache-flink/apache-flink/_build/results?buildId=56603&view=logs&j=3e4dd1a2-fe2f-5e5d-a581-48087e718d53&t=b4612f28-e3b5-5853-8a8b-610ae894217a]
> has the python stage running into a timeout without any obvious reason. The
> [python stage run for
> JDK17|https://dev.azure.com/apache-flink/apache-flink/_build/results?buildId=56603&view=logs&j=b53e1644-5cb4-5a3b-5d48-f523f39bcf06]
> was also getting close to the 4h timeout.
> I'm creating this issue for documentation purposes.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
