I have build tests and have personally tested some Mahout Spark code on my home cluster using Spark 1.0.2 and Hadoop 1.2.1.
Some users are experiencing problems with running this code on their clusters even though they can run it standalone. The only difference is that their Spark master URL is passed in in one case and default local[2] in the other. Is there someone who is using Mahout on a Hadoop 2 + Spark cluster that could run the epinions data through something already in the snapshot-1.0? Two users are having trouble, one is using an AWS hadoop + spark image and the other built their own so I’m a little nervous that there might be a problem somewhere. I can provide data and a command line to run the test. Thanks
