Generate 1 mapper per a file that checks to make sure the keys are sorted within each file. The mapper also generates "$file:begin", first key and "$file:end", last key. The reduce verifies that all of the start/end items are in order. Any output from the reduce is problem report.
To run the program: bin/hadoop jar hadoop-*-examples.jar teravalidate out-dir report-dir
If there is any output, something is wrong and the output of the reduce will have the problem report.