FW: how to run the mapreduce job from localfilesystem but in hadoop cluster(fully distributed mode)

[Roshan Karki <ro...@nepasoft.com.np>]

 

 

From: Roshan Karki 
Sent: Wednesday, November 18, 2009 3:34 PM
To: 'common-user@hadoop.apache.org'
Subject: how to run the mapreduce job from localfilesystem but in hadoop
cluster(fully distributed mode)

 

Hi ,

      I  have setup hadoop in a fully distributed mode(hadoop cluster).
The default file sytem is hdfs.I have two nodes cluster. The map reduce
job works fine when I give inputfile from hdfs location and output is
also generated in hdfs when running WordCount example. My
hadoop-site.xml looks like this:

<?xml version="1.0"?>

<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>

<!-- Put site-specific property overrrides in this file. -->

 

<configuration>

  <property>

    <name>fs.default.name</name>

    <value>hdfs://server1.example.com</value>

  </property>

  <property>

    <name>mapred.job.tracker</name>

    <value>server1.example.com:9001</value>

  </property>

  <property>

    <name>dfs.data.dir</name>

    <value>/home/admin/hadoop/dfs/data</value>

  </property>

  <property>

    <name>dfs.name.dir</name>

    <value>/home/admin/hadoop/dfs/name</value>

  </property>

  <property>

    <name>hadoop.tmp.dir</name>

    <value>/tmp/hadoop</value>

  </property>

  <property>

    <name>mapred.system.dir</name>

    <value>/hadoop/mapred/system</value>

  </property>

  <property>

    <name>dfs.replication</name>

    <value>2</value>

  </property>

</configuration>

 

The problem is that I don't want to give hdfs location for fileinput and
want to generated output in localfilesystem also giving the inputfile
from local but I want to do this in hadoop fully distributed mode
configuration(keeping the hadoop cluster alive).

Giving default filesystem other than hdfs doesn't run cluster.so

Is there any way to achieve this?

I also have nfs server running. All the nodes in hadoop cluster are also
the client of nfs,so they can share the exported /home/admin directory.

Does nfs help in this regard?

Do I have to change my hadoop-site.xml.If so what would it looks like?

Any help will be heartly appreciated.

I am looking forward for the kind response.

 

 

Re: FW: how to run the mapreduce job from localfilesystem but in hadoop cluster(fully distributed mode)

[Steve Loughran <st...@apache.org>]

Roshan Karki wrote:
>  

>  
> 
> The problem is that I don't want to give hdfs location for fileinput and
> want to generated output in localfilesystem also giving the inputfile
> from local but I want to do this in hadoop fully distributed mode
> configuration(keeping the hadoop cluster alive).
> 


Either upload the data files to the HDFS filestore before the run , or 
use file: URLs and make sure that your files are visible on every task 
tracker by mounting the same network drive on the same path, such as 
file:/mount/nfs/server1/work/data