hadoop-hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Vladimir Klimontovich (JIRA)" <j...@apache.org>
Subject [jira] Updated: (HIVE-1419) Policy on deserialization errors
Date Mon, 21 Jun 2010 14:51:27 GMT

     [ https://issues.apache.org/jira/browse/HIVE-1419?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Vladimir Klimontovich updated HIVE-1419:

    Attachment: corrupted_records_0.5_ver2.patch

A bit improved version of patch attached. MapOperator now skips record if deserializer returned

It makes deserializer plugin arch more flexible. If deserializer considers record as non-sense
(corrupted, empty, whatever) it could simply return null and signal Hive not to consider it.

> Policy on deserialization errors
> --------------------------------
>                 Key: HIVE-1419
>                 URL: https://issues.apache.org/jira/browse/HIVE-1419
>             Project: Hadoop Hive
>          Issue Type: Improvement
>          Components: Serializers/Deserializers
>    Affects Versions: 0.5.0
>            Reporter: Vladimir Klimontovich
>            Assignee: Vladimir Klimontovich
>            Priority: Minor
>             Fix For: 0.5.1, 0.6.0
>         Attachments: corrupted_records_0.5.patch, corrupted_records_0.5_ver2.patch, corrupted_records_trunk.patch,
> When deserializer throws an exception the whole map tasks fails (see MapOperator.java
file). It's not always an convenient behavior especially on huge datasets where several corrupted
lines could be a normal practice. Proposed solution:
> 1) Have a counter of corrupted records
> 2) When a counter exceeds a limit (configurable via hive.max.deserializer.errors property,
0 by default) throw an exception. Otherwise just log and exception with WARN level.
> Patches for 0.5 branch and trunk are attached

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message