Getting mapreduce job result from Java API without permanent storage on HDFS

Question

What I need looks like pretty common task but I failed to find 'ready to use' solution.

• I have some Hadoop MapReduce job which produces result as a set of files.
• If it is important, it has several mappers and no reducers but I'd like solution for general case with all of stages.
• I'd like to collect job result and use it just after job execution.
• But I dislike need to store job results permanently on HDFS as with FileOutputFormat so it looks like temporary job storage is option for me.

Any common approach for such cases?

Answer 1

One of solutions for some kind of jobs happened to be counters. It's far from being universal but for some simple cases it can help a lot:

• An Example of Hadoop MapReduce Counter.
• Hadoop definitive guide, counters.