[
https://issues.apache.org/jira/browse/FLINK-40669?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18116024#comment-18116024
]
Martijn Visser commented on FLINK-40669:
----------------------------------------
Another one, nightly 2026-09-16, this time the Hadoop 3.1.3 profile rather than
Java 11, killed at
the 265 minute budget:
https://github.com/apache/flink/actions/runs/35046431137/job/104647064513
All 35 bash e2e tests pass, and the kill lands during the Java e2e phase with
{{org.apache.flink.dist.DynamicParameterITCase}} the only class that never
reported a result:
{code}
[INFO] Running org.apache.flink.dist.DynamicParameterITCase
==========================================================================================
=== WARNING: This task took already 95% of the available time budget of 265
minutes ===
==========================================================================================
{code}
That is FLINK-40070, fixed on master by f9e0756fbfd and not backported, so this
should be closed as
a duplicate once release-2.3 has the backport.
> Java 11 E2E group 2 on release-2.3 is killed at the 234 minute budget while
> the other profiles finish in half the time
> ----------------------------------------------------------------------------------------------------------------------
>
> Key: FLINK-40669
> URL: https://issues.apache.org/jira/browse/FLINK-40669
> Project: Flink
> Issue Type: Bug
> Components: Tests
> Affects Versions: 2.3.0
> Reporter: Martijn Visser
> Priority: Major
> Labels: test-stability
>
> On release-2.3 the {{Java 11 / E2E (group 2)}} leg runs for 309 minutes and
> is killed, while the same group on the other three profiles finishes well
> inside the budget:
> {code}
> === WARNING: This task took already 95% of the available time budget of 234
> minutes ===
> === WARNING: Killing task ===
> ##[error]Process completed with exit code 143.
> {code}
> Durations of the four legs, from the job timestamps:
> {code}
> 2026-09-15 AdaptiveScheduler 133m pass Java 11 309m FAIL Java 21 145m
> pass Hadoop 3.1.3 157m pass
> 2026-09-11 AdaptiveScheduler 148m pass Java 11 309m FAIL Java 21 130m
> pass Hadoop 3.1.3 165m pass
> {code}
> https://github.com/apache/flink/actions/runs/34919695234
> https://github.com/apache/flink/actions/runs/34553046289
> All 35 bash e2e tests pass in that leg and the kill lands after them, during
> the Java e2e phase at {{DynamicParameterITCase}}, so it is the total runtime
> rather than one hung test. The nightlies of 2026-09-12 to 2026-09-14 died
> early on FLINK-40652 and never ran the full suite, which is why this only
> shows in two runs.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
