Dependence of number of reducers on quantity of nodes of a cluster

Question

My hadoop program uses one mapper which breaks input data into a certain number of the portions which number is set in the /usr/countcomputers.txt file (it is read by the mapper function). Further on one portion arrives on each reducer. Thus the number which is set in the /usr/countcomputers.txt file, defines number of reducers. In this regard I had a question: reducers are executed only on computers on which the demon of TaskTracker is launched, or on all nodes, including on what demons of NameNode, by JobTracker and Secondary NameNode are launched? To me it is very important to know the response to this question as the number which is set in the /usr/countcomputers.txt file depends on it, read in the program.

Answer 1

Only a tasktracker is allowed to run reducers (if you configured a number greater than zero for the property mapreduce.tasktracker.reduce.tasks.maximum).

So you can safely assume that only on the hosts where a tasktracker is run, a reduce task will be executed.