[
https://issues.apache.org/jira/browse/TEZ-1894?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jeff Zhang updated TEZ-1894:
----------------------------
Attachment: TEZ-1894-1.patch
Attach patch. Add checkOutputSpec when building MROutput
> No checkOutputSpecs for MROutput
> --------------------------------
>
> Key: TEZ-1894
> URL: https://issues.apache.org/jira/browse/TEZ-1894
> Project: Apache Tez
> Issue Type: Bug
> Reporter: Jeff Zhang
> Assignee: Jeff Zhang
> Priority: Critical
> Attachments: TEZ-1894-1.patch
>
>
> MROutput won't check whether the destination folder exists, so it would cause
> weird result.
> E.g. I run tez WordCount example with 5 partitions, it would generate 5 part
> files, and then run the same WordCount example with 1 partition, it would
> just override one part file, in that case the results of 2 dags coexist in
> the same folder.
> {code}
> Found 6 items
> -rw-r--r-- 1 jzhang supergroup 0 2014-12-28 14:38 output/_SUCCESS
> -rw-r--r-- 1 jzhang supergroup 15 2014-12-28 14:38
> output/part-v001-o000-00000
> -rw-r--r-- 1 jzhang supergroup 8198 2014-12-28 14:37
> output/part-v001-o000-00001
> -rw-r--r-- 1 jzhang supergroup 7372 2014-12-28 14:37
> output/part-v001-o000-00002
> -rw-r--r-- 1 jzhang supergroup 8575 2014-12-28 14:37
> output/part-v001-o000-00003
> -rw-r--r-- 1 jzhang supergroup 6755 2014-12-28 14:37
> output/part-v001-o000-00004
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
