hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Chris Nauroth (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-6208) DataNode caching can leak file descriptors.
Date Tue, 08 Apr 2014 23:36:17 GMT

    [ https://issues.apache.org/jira/browse/HDFS-6208?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13963627#comment-13963627
] 

Chris Nauroth commented on HDFS-6208:
-------------------------------------

There are 2 problems:
# The block file stream and checksum file stream are only closed if there is some kind of
I/O error.  They do not get closed if everything succeeds.  mmap/mlock does not rely on keeping
the source file descriptor open, so we can change our code to close these.
# In test runs, all cached blocks remaining after a test finishes are kept mmap'd into the
process.  We can handle this by explicitly uncaching at shutdown time.

I spotted these issues during runs of {{TestCacheDirectives}} on Windows, where leaked file
descriptors and memory-mapped regions cause subsequent tests to fail, because locks are still
held on the underlying block files in the test data directory.

I have a patch in progress to fix this.

> DataNode caching can leak file descriptors.
> -------------------------------------------
>
>                 Key: HDFS-6208
>                 URL: https://issues.apache.org/jira/browse/HDFS-6208
>             Project: Hadoop HDFS
>          Issue Type: Bug
>          Components: datanode
>    Affects Versions: 2.4.0
>            Reporter: Chris Nauroth
>            Assignee: Chris Nauroth
>
> In the DataNode, management of mmap'd/mlock'd block files can leak file descriptors.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message