While the mapred.input.dir.recursive setting claims to allow recursing all the files within a directory tree being used as an input of a job[1], it doesn't seem to be working for me when using s3 to store the input. Has anyone else seen this and worked around (or had it work, in which case maybe something else is going wrong with my jobs and I'll dig some more)?
Thanks, - Jeremy [1] https://issues.apache.org/jira/browse/MAPREDUCE-1501
