You can specify the HDFS path as follows: FileOutputFormat.setOutputPath(conf, new Path(args[1])); where Path object is of course the location of your output dir.
See this for details http://www.rohitmenon.com/index.php/introducing-mapreduce-part-i/ Regards, Shahab On Thu, Oct 24, 2013 at 11:25 PM, S. Zhou <myx...@yahoo.com> wrote: > Thanks Shahab & Yong. If cluster B (in which I want to dump output) has > url "hdfs://machine.domain:8080" and data folder "/tmp/myfolder", what > should I specify as the output path for MR job? > Thanks > > > On Thursday, October 24, 2013 5:31 PM, java8964 java8964 < > java8...@hotmail.com> wrote: > Just specify the output location using the URI to another cluster. As > long as the network is accessible, you should be fine. > > Yong > > ------------------------------ > Date: Thu, 24 Oct 2013 15:28:27 -0700 > From: myx...@yahoo.com > Subject: Mapreduce outputs to a different cluster? > To: user@hadoop.apache.org > > The scenario is: I run mapreduce job on cluster A (all source data is in > cluster A) but I want the output of the job to cluster B. Is it possible? > If yes, please let me know how to do it. > > Here are some notes of my mapreduce job: > 1. the data source is an HBase table > 2. It only has mapper no reducer. > > Thanks > Senqiang > > > >