In that case, you must configure Hadoop the "normal" way, by editing its
configuration files in conf/. In particular I think this goes in
mapred-site.xml? Because when it is run on a cluster, your local JVM
argumetns would have no way of affecting the cluster.

On Wed, Jun 1, 2011 at 3:25 AM, Qiuyan Xu <qiuyan.xu@mailbox.tu-berlin.de>wrote:

> It doesn't work. I took a look at $MAHOUT_HOME/bin/mahout, MAHOUT_OPTS is
> only used when it's running locally. In my case it's running as
> "$HADOOP_HOME/bin/hadoop" jar $MAHOUT_JOB $CLASS "$@". I don't find a proper
> position to add the compress option.
>
>