Martijn Visser created FLINK-39846:
--------------------------------------

             Summary: IPv6HostnamesITCase runs unconditionally and fails on 
AdaptiveScheduler after JUnit5 migration
                 Key: FLINK-39846
                 URL: https://issues.apache.org/jira/browse/FLINK-39846
             Project: Flink
          Issue Type: Bug
          Components: Tests
            Reporter: Martijn Visser
            Assignee: Martijn Visser


IPv6HostnamesITCase.testClusterWithIPv6host fails in the 
test_cron_adaptive_scheduler tests leg on master, every nightly since 
~2026-04-15. It is a JUnit5-migration regression 
https://issues.apache.org/jira/browse/FLINK-39124

The migration converted the JUnit4 @Rule MiniClusterWithClientResource(... 
.setConfiguration(getConfiguration()) ...) to a static @RegisterExtension 
MiniClusterExtension(...) with a hardcoded configuration, dropping the call to 
getConfiguration(). That method (a) bound the cluster to a local IPv6 address 
and (b) skipped the test (AssumptionViolatedException / IPv6 assumption) when 
no non-loopback IPv6 address is available. Post-migration getConfiguration() is 
dead code, so the test runs unconditionally and without IPv6 configuration, and 
fails under the AdaptiveScheduler because the job uses 
GlobalWindows.createWithEndOfStreamTrigger() (a blocking data exchange the 
AdaptiveScheduler does not support):

{code:java}
Caused by: java.lang.IllegalStateException: The adaptive scheduler supports 
pipelined data exchanges (violated by GlobalWindows ...).
{code}

The CI environment has no usable IPv6 address (confirmed: release-2.0/2.1, 
which still call getConfiguration(), skip the test and stay green; the same CI 
agents run both master and release builds). A secondary effect: on master the 
test no longer applies the IPv6 config at all, so it isn't testing IPv6 
hostnames anymore.

Fix: resolve the IPv6 address once at class load, configure the MiniCluster 
with it when present, self-skip when absent (restoring pre-migration behavior), 
and add @Tag(FailsWithAdaptiveScheduler) so it is also excluded from the 
adaptive profile when IPv6 is present.

Build: 
https://dev.azure.com/apache-flink/apache-flink/_build/results?buildId=75618



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to