[
https://issues.apache.org/jira/browse/KAFKA-17371?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17875170#comment-17875170
]
Ao Li edited comment on KAFKA-17371 at 8/20/24 12:35 PM:
---------------------------------------------------------
[~frankvicky] Thanks for taking care of this! Making both
`DefaultTaskExecutor::currentTask` and `DefaultTaskExecutor::runOnce`
synchronized is a quick solution. However, I'm not sure if this will introduce
any performance regression.
[~chia7712] I'm currently running a concurrency testing tool that reruns all
Kafka tests with different thread schedules. This helped me to find many bugs.
While I can reproduce these failures and (sometimes) identify the root cause, I
don't have enough context to fix these issues since I am not familiar with the
codebase. Please let me know if you want me to report these bugs through
another channel.
was (Author: JIRAUSER306156):
[~frankvicky] Thanks for taking care of this! Making both
`DefaultTaskExecutor::currentTask` and `DefaultTaskExecutor::runOnce`
synchronized a quick solution. However, I'm not sure if this will introduce any
performance regression.
[~chia7712] I'm currently running a concurrency testing tool that reruns all
Kafka tests with different thread schedules. This helped me to find many bugs.
While I can reproduce these failures and (sometimes) identify the root cause, I
don't have enough context to fix these issues since I am not familiar with the
codebase. Please let me know if you want me to report these bugs through
another channel.
> Flaky test in DefaultTaskExecutorTest.shouldUnassignTaskWhenRequired
> --------------------------------------------------------------------
>
> Key: KAFKA-17371
> URL: https://issues.apache.org/jira/browse/KAFKA-17371
> Project: Kafka
> Issue Type: Bug
> Reporter: Ao Li
> Assignee: TengYao Chi
> Priority: Minor
>
> Please see this fork https://github.com/aoli-al/kafka/tree/KAFKA-251 for a
> deterministic reproduction.
> The test failed with
> {code}
> expected: not <null>
> org.opentest4j.AssertionFailedError: expected: not <null>
> at
> org.junit.jupiter.api.AssertionFailureBuilder.build(AssertionFailureBuilder.java:152)
> at
> org.junit.jupiter.api.AssertionFailureBuilder.buildAndThrow(AssertionFailureBuilder.java:132)
> at org.junit.jupiter.api.AssertNotNull.failNull(AssertNotNull.java:49)
> at
> org.junit.jupiter.api.AssertNotNull.assertNotNull(AssertNotNull.java:35)
> at
> org.junit.jupiter.api.AssertNotNull.assertNotNull(AssertNotNull.java:30)
> at org.junit.jupiter.api.Assertions.assertNotNull(Assertions.java:304)
> at
> org.apache.kafka.streams.processor.internals.tasks.DefaultTaskExecutorTest.shouldUnassignTaskWhenRequired(DefaultTaskExecutorTest.java:233)
> at java.base/java.lang.reflect.Method.invoke(Method.java:580)
> at java.base/java.util.ArrayList.forEach(ArrayList.java:1596)
> at java.base/java.util.ArrayList.forEach(ArrayList.java:1596)
> {code}
> The root cause of the failure is that `currentTask =
> taskManager.assignNextTask(DefaultTaskExecutor.this);` is not an atomic
> operation. This means that calling `taskManager.assignNextTask` will unblock
> the `verify(taskManager,
> timeout(VERIFICATION_TIMEOUT)).assignNextTask(taskExecutor);` statement in
> the test method.
> If `assertNotNull(taskExecutor.currentTask());` is executed before the
> assignment `currentTaks = [...]` the test will fail.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
