How to read Json Data from online file?

If a process is reading data from a file on HDFS named fileA, then another process moves a file named fileA over the top of it, will the read of the original file break?

• Here is what I think will happen: Say the original fileA is comprised of blocks 102, 103, and 104. When the read is initiated, fileA will be dereferenced to those blocks, and those blocks will be read. While the read is happening, fileA will be redirected (by the move operation) to point to, say, blocks 502, 503 and 504. But the old blocks are still around, and they will be read. But what if another operation is also writing to the cluster? What if it overwrites blocks 100, 101, 102, and 103? Or does HDFS remember that someone is reading those blocks, and not allocate them to hold the new data being written?

• Answer:

  No. New file is copied as a duplicate. You can avoid that in your load process though.