I haven't been able to try this with distcp, but a similar thing happens with hdfs dfs commands. What I found is if you have your target folder created (e.g. hdfs dfs -mkdir /e/f/), then copying into that folder will give you all of your CSVs as separate files. If you don't have /e/f/ created ahead of time, then Hadoop will create it for you and rename your source csv to be called "f". Hope that makes sense and helps.
... View more
