Converting to.dfs argument to keyval with a NULL key about rmr2

[piccolbo]

Please provide a test case. If I can't repro, by and large, I can't fix it.

[piccolbo]

Also check this
http://stackoverflow.com/questions/11898663/what-is-the-possible-reasons-of-log4j-error-could-not-find-value-for-key-log4j

I suspect the first part of the warning is pretty harmless. We added those warnings recently as an aid to people porting programs from 1.x to 2.x As you said in a separate message, providing a non-NULL key is a enough to prevent the error, hence they seem related. But I suspect any warning would trigger the writing of a log entry and trigger the log4j issue. To test that you just need to take a mapreduce job that completes successfully and add a warning("testing log4j theory") to the map function and see what happens. Thanks