You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Prasanth Jayachandran (JIRA)" <ji...@apache.org> on 2016/06/06 18:58:21 UTC

[jira] [Commented] (HIVE-13954) Parquet logs should go to STDERR

    [ https://issues.apache.org/jira/browse/HIVE-13954?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15316994#comment-15316994 ] 

Prasanth Jayachandran commented on HIVE-13954:
----------------------------------------------

This writes to $TMPDIR/parquet-%u.log where %u is unique number to resolve conflicts. The default location for hive log is $TMPDIR/$USER/hive.log. We cannot use the same file for logging because with java file logging we cannot access the user name in properties file, hence the above pattern for parquet log file. 

> Parquet logs should go to STDERR
> --------------------------------
>
>                 Key: HIVE-13954
>                 URL: https://issues.apache.org/jira/browse/HIVE-13954
>             Project: Hive
>          Issue Type: Bug
>    Affects Versions: 2.2.0
>            Reporter: Prasanth Jayachandran
>            Assignee: Prasanth Jayachandran
>         Attachments: HIVE-13954.1.patch
>
>
> Parquet uses java util logging. When java logging is not configured using default logging.properties file, parquet's default fallback handler writes to STDOUT at INFO level. Hive writes all logging to STDERR and writes only the query output to STDOUT. Writing logs to STDOUT may cause issues when comparing query results. 
> If we provide default logging.properties for parquet then we can configure it to write to file or stderr.



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