for maximum integration it would be nice to have that a mapreduce call, on exit, clean up the related hadoop job. This would happen only if mapreduce terminates abnormally, for instance gets interrupted. This should work exactly the same at the prompt or in a script. It's not clear what the implementation approach could be, other than a) the kill command is included in the job output so capturing that (capture.otuput?) and calling system is one way of doing it b)the on.exit function is the way to add this clean up action
This is still not a plan because the information on how to kill a job is available after on exit can be called
Went through a number of experiments today using mostly tryCatch and I can confirm the statement that the output can be captured only if the job completes. If anyone has ideas I am listening.