On Thu, Mar 4, 2010 at 7:41 AM, Robin Anil <robin.a...@gmail.com> wrote:
> Based on what i have in mind, the usage will just be > > mahout vectorize -i s3://input -o s3://output -tmp hdfs://file (here, there > is a risk of fixing a exact path and not knowing the hadoop user, I would > have preferred a relative path) > So according to Peter and his support devs over at Amazon, EMR is already set to take any unqualified Path and have the default FileSystem to be (local) HDFS, so we won't need to do anything special here - if we want to have temp directories, we just specify them in the usual unqualified way, and they'll hit the local HDFS, and then if you want to grab stuff in and out of s3, you specify directly: mahout vectorize -i s3://input -o s3://output -tmp tmp/myJobsTemp -jake
