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 2016/08/22 08:58:20 UTC

[jira] [Commented] (HADOOP-13531) S3A output streams to share a single LocalDirAllocator for round-robin drive use

    [ https://issues.apache.org/jira/browse/HADOOP-13531?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15430351#comment-15430351 ] 

Steve Loughran commented on HADOOP-13531:
-----------------------------------------

May need to add a new unit test for this if we want to verify round-robin use; otherwise, regression testing.

> S3A output streams to share a single LocalDirAllocator for round-robin drive use
> --------------------------------------------------------------------------------
>
>                 Key: HADOOP-13531
>                 URL: https://issues.apache.org/jira/browse/HADOOP-13531
>             Project: Hadoop Common
>          Issue Type: Sub-task
>          Components: fs/s3
>    Affects Versions: 2.8.0
>            Reporter: Steve Loughran
>            Priority: Minor
>
> {{S3AOutputStream}} uses {{LocalDirAllocator}} to choose a directory from the comma-separated list of buffers —but it creates a new instance for every output stream. This misses a key point of the allocator: for it to do round-robin allocation, it needs to remember the last disk written to. If a new instance is used for every file: no history.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-issues-help@hadoop.apache.org