Moving files to HDFS or any file system, if implemented, involves changing the namespace and not moving the actual data. Looking through the code, only changes in the namespace (memory and editing log) in the node name are made.
From NameNode.java class
- NameNode manages two critical tables:
- 1) filename-> blocksequence (namespace)
- 2) block-> machine translation ("inodes")
Only the first part needs to be changed; there is no need to block the list of machines. I have not tried, but I think everything should be OK.
source share