Not that I know of, the block allocations are done by the NameNode for several reasons/edge cases, and that is why the write path is as it is today.
What problem are you trying to solve though, that led you to ask this? If we have more info about the problem to solve, we can provide better answers. On Thu, Jul 12, 2012 at 2:37 AM, Grandl Robert <rgra...@yahoo.com> wrote: > Hi, > > It is possible to write to a HDFS datanode w/o relying on Namenode, i.e. to > find the location of Datanodes from somewhere else ? > > Thanks, > Robert -- Harsh J
