[
https://issues.apache.org/jira/browse/IGNITE-28375?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18069832#comment-18069832
]
Maksim Davydov commented on IGNITE-28375:
-----------------------------------------
https://ci2.ignite.apache.org/buildConfiguration/IgniteTests24Java8_PlatformNetWindows?branch=pull%2F12944%2Fhead#all-projects
Ran the .NET (Windows) suite a few times - no flakes so far.
> Fix flaky JniThreadDetachTest
> -----------------------------
>
> Key: IGNITE-28375
> URL: https://issues.apache.org/jira/browse/IGNITE-28375
> Project: Ignite
> Issue Type: Task
> Reporter: Maksim Davydov
> Assignee: Maksim Davydov
> Priority: Minor
> Time Spent: 20m
> Remaining Estimate: 0h
>
> JniThreadDetachTest in the .NET module verifies that using Ignite APIs from
> CLR threads does not leak JVM threads, ensuring proper JNI thread
> attach/detach behavior.
> The test works by capturing a snapshot of JVM thread names before and after
> executing multi-threaded cache operations from CLR threads, and asserting
> that no additional threads remain after execution.
> Currently, the test is flaky due to the appearance of internal Ignite threads
> with the 'mgmt-' prefix in the thread snapshot obtained via
> _PlatformThreadNamesTask_. These threads belong to Ignite’s internal
> management/statistics thread pool and may be lazily initialized, meaning they
> can appear nondeterministically depending on timing and execution paths.
> As a result, these mgmt- threads are not related to JNI leaks and should be
> excluded from comparison to avoid false positives.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
