[
https://issues.apache.org/jira/browse/KAFKA-9195?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16975623#comment-16975623
]
Matthias J. Sax commented on KAFKA-9195:
----------------------------------------
Could we call `KafkaStreams#cleanup()` before we start the application to make
sure the is no dirty state? Might be simpler than randomizing the state
directory? Or maybe use different application.ids (even if this has this
drawback, that we might get overlapping id if we are not careful.)
> Use Randomized State Directory Names in Streams System Tests
> -------------------------------------------------------------
>
> Key: KAFKA-9195
> URL: https://issues.apache.org/jira/browse/KAFKA-9195
> Project: Kafka
> Issue Type: Improvement
> Components: streams, system tests
> Reporter: Bruno Cadonna
> Priority: Major
>
> Currently, the state directory property in streams' system tests is set to
> the {{PERSISTENT_ROOT}} variable. Since Streams applications in different
> tests have the same application ID and the state directory path consists of
> state directory property + application ID + task ID, it might happen that a
> dirty state directory of one test is re-used by another test if the state
> directory is not properly cleaned up. This may lead to unexpected results and
> false positive and/or flaky failures.
> The state directory property shall be set to a randomized path inside
> {{PERSISTENT_ROOT}} to ensure that tests may not interfere with each other in
> the case of missing state clean-ups.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
