[ 
https://issues.apache.org/jira/browse/HADOOP-5657?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Chris Douglas updated HADOOP-5657:
----------------------------------

    Attachment: 5657-0.patch

The following are now validated in the reduce:
* Each map produces one record for each of 4096 small keys
* Includes unique large records, each straddled by a pair of small records from 
another map (to detect corruption from the merge)
* Changes some parameters for {{testReduceFromDisk}} to make intermediate 
merges with in-memory data occur occasionally

> Validate data passed through TestReduceFetch
> --------------------------------------------
>
>                 Key: HADOOP-5657
>                 URL: https://issues.apache.org/jira/browse/HADOOP-5657
>             Project: Hadoop Core
>          Issue Type: Improvement
>          Components: mapred, test
>            Reporter: Chris Douglas
>         Attachments: 5657-0.patch
>
>
> While TestReduceFetch verifies the reduce semantics for reducing from 
> in-memory segments, it does not validate the data it reads. Data corrupted 
> during the merge will not be detected.

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

Reply via email to