问题
During the mapreduce job, the individual tasks will be running on random nodes, Is there any way to restrict the nodes in which the tasks should run ?
回答1:
Hadoop doesn't pick the nodes to run the tasks on random. Data locality is considered or else there would be a lot of network overhead.
There is no affinity between the tasks and the nodes on which they run. Hadoop doesn't offer any such functionality.
来源：https://stackoverflow.com/questions/15970044/how-to-specify-individual-tasktrackersnodes-on-which-a-mapreduce-should-run