I tried to make a my counter to check the operation time of Maptask.run and ReduceTask.run using global counter or even LOG.info().
In a pseudo-distributed mode, using builded mapreduce-core-snapshot.jar which I modified in the MapTask.java and ReduceTask.java to make my goal, it works perfectly as I expected. However, in a real cluster, the source code I injected isn't work and just system-default counter variables are showing. I think there are different logging logic or security issue between pseudo-distributed and cluster mode, but I am not sure about that. Is there anyone who can let me know about it? Thanks!
