The client splits the file. The block size attribute is a per-file one, sent along with the file creation request. Blocks are requested as-it-goes, not pre-allocated.
On Sun, Mar 31, 2013 at 2:15 PM, Sai Sai <[email protected]> wrote: > Here is my understanding about putting a file into hdfs: > A client contacts name node and gets the location of blocks where it needs > to put the blocks in data nodes. > But before this how does the name node know how many blocks it needs to > split a file into. > Who splits the file is it the client itself or is it the name node. > Since the file is never written to the name node so there is no chance it is > the name node. > Please clarify. > Thanks > Sai > -- Harsh J
