You are viewing a plain text version of this content. The canonical link for it is here.

Posted to common-issues@hadoop.apache.org by "Steve Loughran (JIRA)" <ji...@apache.org> on 2014/01/30 10:44:09 UTC

[jira] [Commented] (HADOOP-10309) S3 block filesystem should more aggressively delete temporary files

    [ https://issues.apache.org/jira/browse/HADOOP-10309?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13886448#comment-13886448 ] 

Steve Loughran commented on HADOOP-10309:
-----------------------------------------

submitting patch, though as jenkins doesn't run the S3 tests it'll need a manual run through. 

> S3 block filesystem should more aggressively delete temporary files
> -------------------------------------------------------------------
>
>                 Key: HADOOP-10309
>                 URL: https://issues.apache.org/jira/browse/HADOOP-10309
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: fs/s3
>            Reporter: Joe Kelley
>            Priority: Minor
>         Attachments: HADOOP-10309.patch
>
>
> The S3 FileSystem reading implementation downloads block files into a configurable temporary directory. deleteOnExit() is called on these files, so they are deleted when the JVM exits.
> However, JVM reuse can lead to JVMs that stick around for a very long time. This can cause these temporary files to build up indefinitely and, in the worst case, fill up the local directory.
> After a block file has been read, there is no reason to keep it around. It should be deleted.
> Writing to the S3 FileSystem already has this behavior; after a temporary block file is written and uploaded to S3, it is deleted immediately; there is no need to wait for the JVM to exit.



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)