hadoop-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Piter85 Piter85 <piter1...@hotmail.it>
Subject RE: How get information messages when a JobControl is used ?
Date Wed, 12 Sep 2012 13:06:20 GMT

Ok I will do it!

From: Rekha_Joshi@intuit.com
To: user@hadoop.apache.org
Subject: Re: How get information messages when a JobControl is used ?
Date: Wed, 12 Sep 2012 11:41:21 +0000







Good
 that web hdfs is sufficient for now, Piter!
The
 counters are part of o.a.h.mapreduce.Job so you can get them as job.getCounters()..etc
 or via JobInProgress. It is not a JobControl feature as such, so they will not be directly
in JobControl/ControlledJob API.



However
Bertrand's point is an important one, if there are identified synchronization, concurrency
issues on JobControl, the values on webhdfs too will reflect that ..unless as the collection
of counters is happening within context..hmm..well..

Do please keep us updated if the values you see seem incorrect!



Thanks
Rekha





From: Piter85 Piter85 <piter1485@hotmail.it>

Reply-To: <user@hadoop.apache.org>

Date: Wed, 12 Sep 2012 13:12:23 +0200

To: <user@hadoop.apache.org>

Subject: RE: How get information messages when a JobControl is used ?







Hi Rekha and Bertrand! Thanks for the answers! Ok I see that in web interface (_logs->history->job_.....)
there are infos about executions of jobs.

I hope that this infos will be enough for me.



As I said before, scanning APIs, the only method that I found was ControlledJob:toString().



Bye! :)



Piter






Date: Wed, 12 Sep 2012 12:21:24 +0200

Subject: Re: How get information messages when a JobControl is used ?

From: dechouxb@gmail.com

To: user@hadoop.apache.org



But as far as I know there is no way to have a snapshot of the JobControl state.

https://issues.apache.org/jira/browse/MAPREDUCE-3562



I was trying only to get the state of all jobs and it is not possible to get a consistent
view.

For Map/Reduce progress, I guess you could the same by digging into the APIs.

But I am afraid you will have the same problems.



Regards



Bertrand



On Wed, Sep 12, 2012 at 12:09 PM, Joshi, Rekha 
<Rekha_Joshi@intuit.com> wrote:



Hi Piter,



JobControl just means there are multiple complex jobs, but you will see the information for
each job on your hadoop web interface webhdfs still, wouldn't you?
Or if that does not work, you might need to use Reporters/Counters to get the log info data
in custom format as needed.



Thanks
Rekha








From: Piter85 Piter85 <piter1485@hotmail.it>

Reply-To: <user@hadoop.apache.org>

Date: Wed, 12 Sep 2012 11:34:27 +0200

To: <user@hadoop.apache.org>

Subject: How get information messages when a JobControl is used ?







Hi! I'm using JobControl (v. 1.0.3)  to chain two MapReduce applications. It works and creates
output data, but it doesn't give me back information messages as number of mappers, number
of records in input or in output, etc...



It only returns messages like this :

12/09/12 09:56:38 WARN mapred.JobClient: Use GenericOptionsParser for parsing the arguments.
Applications should implement Tool for the same.

12/09/12 09:56:38 INFO input.FileInputFormat: Total input paths to process : 4



I tried to use ControlledJob's toString() method but it returns me only this kind of message:



job name:    songsort

job id:    jobctrl1

job state:    RUNNING

job mapred id:    job_201209120942_0005

job message:    just initialized

job has 1 dependeng jobs:

     depending job 0:    canzoni



Any idea to get back remainder infos?



Bye!












-- 

Bertrand Dechoux





 		 	   		  
Mime
View raw message