[
https://issues.apache.org/jira/browse/YARN-5939?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15708143#comment-15708143
]
Weiwei Yang commented on YARN-5939:
-----------------------------------
Submitting a patch to fix this. It is found that in {{FSDownload}} class, there
are quite of some places leaks FileSystem resources. I improved
{{TestFSDownload}} and added {{verifyFileSystemClosed}} at the end of each test
to verify file system is properly closed. Without the patch, I saw
{noformat}
TestFSDownload.testDirDownload:743->verifyFileSystemClosed:784 FileSystem was
not closed expected:<0> but was:<16>
TestFSDownload.testUniqueDestinationPath:780->verifyFileSystemClosed:784
FileSystem was not closed expected:<0> but was:<3>
TestFSDownload.testDownloadArchiveZipWithTurkishLocale:617->verifyFileSystemClosed:784
FileSystem was not closed expected:<0> but was:<3>
TestFSDownload.testDownloadPatternJar:592->verifyFileSystemClosed:784
FileSystem was not closed expected:<0> but was:<3>
TestFSDownload.testDownloadArchive:584->verifyFileSystemClosed:784 FileSystem
was not closed expected:<0> but was:<3>
TestFSDownload.testDownloadArchiveTgz:625->verifyFileSystemClosed:784
FileSystem was not closed expected:<0> but was:<3>
TestFSDownload.testDownloadArchiveZip:600->verifyFileSystemClosed:784
FileSystem was not closed expected:<0> but was:<3>
TestFSDownload.testDownloadBadPublic:340->verifyFileSystemClosed:784 FileSystem
was not closed expected:<0> but was:<1>
TestFSDownload.testDownloadPublicWithStatCache:423->verifyFileSystemClosed:784
FileSystem was not closed expected:<0> but was:<10>
TestFSDownload.testDownload:500->verifyFileSystemClosed:784 FileSystem was not
closed expected:<0> but was:<30>
{noformat}
after applied the patch, these issues are fixed. Hi [~lxw668899], could you
please check if the fix would resolve the problem ?
> Disabling the cache of FileSystem leads to case that never close the
> FileSystem
> -------------------------------------------------------------------------------
>
> Key: YARN-5939
> URL: https://issues.apache.org/jira/browse/YARN-5939
> Project: Hadoop YARN
> Issue Type: Bug
> Components: nodemanager
> Affects Versions: 2.5.1, 2.7.3
> Reporter: liuxiangwei
> Assignee: Weiwei Yang
> Attachments: YARN-5939.01.patch
>
> Original Estimate: 24h
> Remaining Estimate: 24h
>
> Background
> To use our self-defined FileSystem class, the item of configuration
> "fs.%s.impl.disable.cache" should set to true.
> In YARN's source code, the class named
> "org.apache.hadoop.yarn.util.FSDownload" use getFileSystem but never close,
> which leading to file descriptor leak because our self-defined FileSystem
> class close the file descriptor when the close function is invoked.
> My Question below:
> 1. whether invoking "getFileSystem" but never close is YARN's expected
> behavior
> 2. what should we do in our self-defined FileSystem resolve it.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
