hadoop-hdfs-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Eduardo Afonso Ferreira <eafon...@yahoo.com>
Subject Re: no _SUCCESS file in MR output directory.
Date Mon, 06 May 2013 14:24:34 GMT
I'm not sure if Pig disables it or not, but I remember I had issues when that file was to be
created by the MR jobs due to the fact of Oozie or Pig removing temporary directories or something
like that. I remember seeing an exception about failure to create the SUCCESS file, so I started
using the following property in my workflow pig action to disable that:

        <pig>
            ...
            <configuration>

                <property>
                    <name>mapreduce.fileoutputcommitter.marksuccessfuljobs</name>
                    <value>false</value>
                </property>
                ...
            </configuration>
            ...
        </pig>



________________________________
 From: Rahul Bhattacharjee <rahul.rec.dgp@gmail.com>
To: "user@hadoop.apache.org" <user@hadoop.apache.org>; user@oozie.apache.org 
Sent: Monday, May 6, 2013 3:48 AM
Subject: Re: no _SUCCESS file in MR output directory.
 

I wanted to confirm whether oozie disables the _SUCCESS file creation when
it triggers a MR job.

I am triggering a MR job (actually a bunch of 'em) from oozie and the
workflow completes successfully , however I do not see any kind of _SUCCESS
file in the output directory.
When I set the file output committer's configuration
(mapreduce.fileoutputcommitter.
marksuccessfuljobs
) to true.It generates the success file. Wanted to confirm if oozie does
the disabling of success file creation.

​Thanks,
Rahul​


On Mon, May 6, 2013 at 12:34 PM, Rahul Bhattacharjee <
rahul.rec.dgp@gmail.com> wrote:

> Oozie is being used for triggering the MR job. Looks like oozie disables
> the success file creation using the configuration that you have mentioned
> for FileOutputCommitter.
>
> I have enabled it by setting this property in conf.
>
> Rahul
>
>
> On Mon, May 6, 2013 at 9:38 AM, Rahul Bhattacharjee <
> rahul.rec.dgp@gmail.com> wrote:
>
>> Thanks Harsh for the pointers. I will find out more on this.
>>
>>
>> On Sun, May 5, 2013 at 11:26 PM, Harsh J <harsh@cloudera.com> wrote:
>>
>>> I can think of a few, most obvious ones:
>>>
>>> 1. Job didn't succeed and/or the file was deleted (*shields self*)
>>> 2. Job overrode the default FileOutputCommitter with something that
>>> doesn't do success marking.
>>> 3. Job specifically asked to not create such files, via config
>>> mapreduce.fileoutputcommitter.marksuccessfuljobs or so, set to false.
>>>
>>> On Sun, May 5, 2013 at 9:54 PM, Rahul Bhattacharjee
>>> <rahul.rec.dgp@gmail.com> wrote:
>>> > Hi,
>>> >
>>> >
>>> > A few days back , I was going through a MR job's output , but there
>>> wasn't
>>> > any _SUCCESS file in the output directory.
>>> > I was wondering what all reasons for this  (no _SUCCESS file)?
>>> >
>>> > Thanks,
>>> > Rahul
>>>
>>>
>>>
>>> --
>>> Harsh J
>>>
>>
>>
>
Mime
View raw message