Hey! I think Ufuk is onto the right problem. Form the logs it seems as if the master can access the files, but the workers cannot.
You can always get the result in a single file, by setting the parallelism of the sink task to one, for example line "result.writeAsText(path).parallelism(1)". The counting reducers will run in parallel, but ship their results to a single file writer. We are planning to add support to move program results back to the master/client, but it will be a few weeks. Of course, this is only feasibly if the results are not huge. Greetings, Stephan
