[
https://issues.apache.org/jira/browse/HADOOP-19985?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jose Luis López updated HADOOP-19985:
-------------------------------------
Description:
h3. Problem
{\{TestMultipleArchiveFiles}} is excluded in
{\{.github/gha-tests/exclude-tests.txt}} and the reason is not recorded.
HADOOP-19963 added a \{{tearDown}} that shuts down the class's MiniDFSCluster
and
MiniMRCluster, but that does not explain the exclusion. The class declares no
tests of its own and inherits exactly one, \{{TestStreaming.testCommandLine}}.
With \{{reuseForks=false}} the fork runs that single test and exits, so the
leaked clusters lived for seconds inside a JVM that was about to die. Nothing
accumulated.
Whatever made this class fail or abort under GHA is therefore still unknown and
still present.
h3. Scope
Determine the actual failure mode before proposing any removal. The class stands
up a full MiniDFSCluster plus MiniMRCluster in its *constructor*, which runs
before any \{{@BeforeEach}} and outside the reach of \{{tearDown}} if it throws;
that is the first thing to rule in or out. Its excluded siblings
TestMultipleCachefiles, TestSymLink and TestFileArgs are worth checking for the
same pattern - TestFileArgs shares the TestStreaming base.
was:pending
> Re-enable TestMultipleArchiveFiles in GHA
> -----------------------------------------
>
> Key: HADOOP-19985
> URL: https://issues.apache.org/jira/browse/HADOOP-19985
> Project: Hadoop Common
> Issue Type: Sub-task
> Components: common, test
> Reporter: Jose Luis López
> Priority: Major
>
> h3. Problem
> {\{TestMultipleArchiveFiles}} is excluded in
> {\{.github/gha-tests/exclude-tests.txt}} and the reason is not recorded.
> HADOOP-19963 added a \{{tearDown}} that shuts down the class's MiniDFSCluster
> and
> MiniMRCluster, but that does not explain the exclusion. The class declares no
> tests of its own and inherits exactly one, \{{TestStreaming.testCommandLine}}.
> With \{{reuseForks=false}} the fork runs that single test and exits, so the
> leaked clusters lived for seconds inside a JVM that was about to die. Nothing
> accumulated.
> Whatever made this class fail or abort under GHA is therefore still unknown
> and
> still present.
> h3. Scope
> Determine the actual failure mode before proposing any removal. The class
> stands
> up a full MiniDFSCluster plus MiniMRCluster in its *constructor*, which runs
> before any \{{@BeforeEach}} and outside the reach of \{{tearDown}} if it
> throws;
> that is the first thing to rule in or out. Its excluded siblings
> TestMultipleCachefiles, TestSymLink and TestFileArgs are worth checking for
> the
> same pattern - TestFileArgs shares the TestStreaming base.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
