You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@sqoop.apache.org by "Richard (JIRA)" <ji...@apache.org> on 2014/08/26 10:53:59 UTC

[jira] [Commented] (SQOOP-1392) Create the temporary directory inside task working dir rather then in tmp

    [ https://issues.apache.org/jira/browse/SQOOP-1392?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14110509#comment-14110509 ] 

Richard commented on SQOOP-1392:
--------------------------------

For _Avro_, _Text_, _Sequence_ file, *FileOutputFormat.getWorkOutputPath(context)* is used to get temp work output path, since the *OutputFormat* for these three types is *FileOutputFormat*.
However, *DatasetKeyOutputFormat* extends *OutputFormat* directly instead of *FileOutputFormat*. As a result, *FileOutputFormat.getWorkOutputPath(context)* could not be used in _Parquet_ file.
There are 2 solutions to resolve this.
1. Make *DatasetKeyOutputFormat* extends *FileOutputFormat*. Is there any special reason not to implement *DatasetKeyOutputFormat*?
2. Generate temp work output path manually, like *patch.diff* attached. It could make less impact to Kite SDK and Sqoop.

> Create the temporary directory inside task working dir rather then in tmp
> -------------------------------------------------------------------------
>
>                 Key: SQOOP-1392
>                 URL: https://issues.apache.org/jira/browse/SQOOP-1392
>             Project: Sqoop
>          Issue Type: Improvement
>          Components: tools
>            Reporter: Qian Xu
>            Priority: Minor
>         Attachments: patch.diff
>
>
> Several mappers will create a temporary directory in /tmp ranther then inside the task working directory. We should change the pattern.



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