I am running Hadoop jobs on a cluster whose jobtracker and
file system are on a machine in my network called mycluster.

 conf.set("fs.default.name","hdfs://mycluster:9000");
 conf.set("mapred.job.tracker","mycluster:9001);

 If I set in the configuration as shown , set the configuration in a Tool
and call run
the Tool run command, the job runs on the cluster.

My problem is that even when in the configuration I say
conf.set("user.name","MyDesiredUser");

the job runs as the local user.

On my cluster the files and directories are created as the local user
StevesPc\Steve.
In my cluster hdfs-site.xml has the property

<property>
   <name>dfs.permissions</name>
   <value>false</value>
   <final>true</final>
</property>

and in hadoop-policy.xml
all properies are set to * as shown below
 <property>
    <name>security.client.protocol.acl</name>
    <value>*</value>
  </property>

On my cluster things work but on other clusters they do not. The
documentation is not very
clear on permissions and I would like a good discussion on this issue.
Ideally I would want
to set the user (and provide a password) for a hadoop job but still run on
the local box

--
Steven M. Lewis PhD
4221 105th Ave NE
Kirkland, WA 98033
206-384-1340 (cell)
Skype lordjoe_com