Allocator does not initialize lotSize during hdfs mover process

Currently, when we execute 

hdfs mover -p /some/path

the moverThreadAllocator in org.apache.hadoop.hdfs.server.balancer.Dispatcher does not initialize lotSize according to  dfs.mover.moverThreads and dfs.datanode.balance.max.concurrent.moves.  So, when we invoke moverThreadAllocator.allocate() method, it will always return 1.