-Re: Getting job information, joining a job in hadoop 0.20.1
Saptarshi Guha 2010-05-26, 18:48
I found give, counters as obtained from RunningJob, I can still use it with
the new style counters.
But I'm still wondering how to get a RunningJob from a JobId.
Is my Hadoop old?
On Wed, May 26, 2010 at 2:27 PM, Saptarshi Guha <[EMAIL PROTECTED]>wrote:
> I'm using Hadoop 0.20.1, I submitted a job using the
> org.apache.hadoop.mapreduce.Job approach e.g.
> org.apache.hadoop.mapreduce.Job _job
> However, I would like to,from another Java program, monitor this job:
> 1) get map/reduce progress
> 2) get counters
> 3) status
> 4) and the ability to join i.e wait for completion.
> say jd = "job_201003190003_1855"
> org.apache.hadoop.mapreduce.JobID jid > org.apache.hadoop.mapreduce.JobID.forName(jd);
> org.apache.hadoop.mapred.JobClient jclient = new
> Now, what?
> says, I should use
> 1. public RunningJob getJob(JobID jobid)
> but, in my installation of hadoop 0.20.1, i do not have such a method.
> 2. Even if I could, runningjob.getCounters() returns the old deprecated
> Counters. Is this compatible with the new mapreduce code?
> Is there any org.apache.mapreduce code that given the JobId, I can monitor
> and wait for completion for the job?
> I can get a org.apache.hadoop.mapreduce.JobContext with the Job ID, but
> there should be a way to get the Job.
> One option, is to serialize the job_ variable and load it in later(not my
> first choice)
> State: RUNNING
> Started: Fri Mar 19 00:03:40 EDT 2010
> Version: 0.20.1, r810220
> Compiled: Tue Sep 1 20:55:56 UTC 2009 by oom
> Identifier: 201003190003