[
https://issues.apache.org/jira/browse/FLINK-39723?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aleksandr Savonin updated FLINK-39723:
--------------------------------------
Description:
KafkaTableITCase.testExactlyOnceSink is flaky due to a topic metadata
propagation race in createTestTopic
Observed in: CI run
[https://github.com/apache/flink-connector-kafka/actions/runs/26227821749]
was:
KafkaTableITCase.testExactlyOnceSink intermittently fails in CI with
ProducerFencedException caused by transactional producer state leaking between
test runs on the same Kafka broker instance.
Observed in: CI run
https://github.com/apache/flink-connector-kafka/actions/runs/26227821749
Ideas/potential fixes to think about:
- Use randomized/unique transactional ID prefixes per test run (similar to how
topic names already use random UUIDs)
- Or explicitly abort any open transactions with the same ID before the test
starts
- Or ensure the Kafka container is fully restarted between parameterized
variants that share transactional ID space
> KafkaTableITCase.testExactlyOnceSink fails due to transactional ID reuse
> across test runs
> -----------------------------------------------------------------------------------------
>
> Key: FLINK-39723
> URL: https://issues.apache.org/jira/browse/FLINK-39723
> Project: Flink
> Issue Type: Bug
> Components: Connectors / Kafka
> Reporter: Aleksandr Savonin
> Assignee: Aleksandr Savonin
> Priority: Major
> Labels: pull-request-available
> Fix For: kafka-5.1.0
>
>
> KafkaTableITCase.testExactlyOnceSink is flaky due to a topic metadata
> propagation race in createTestTopic
> Observed in: CI run
> [https://github.com/apache/flink-connector-kafka/actions/runs/26227821749]
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
