Hi Hadoop folks: I have a 8 nodes cluster and use copyFromLocal to write some text workload into dfs. I tried to write it on different nodes in the same cluster. That is ,I log on to each machine in the same cluster after hadoop boots. In each machine, I use: bin/hadoop dfs -copyFromLocal workload.txt. I was expecting in this way, each node's local disk will be reduce because its contributed disk space for dfs is consumed. However, after trying a few times, I found that only the disk space of namenode is consumed whether or not copyFromLocal is doing from namenode machine or data node machines. Thus, it seems to have a bottleneck on the total bytes that can write in dfs. Is this a design setting of hadoop or there some ways to let the workload spread in non-name node's disk? Thanks. Richard
