How does Hadoop s block replacement policy work

Question

I'm working on a Hadoop cluster and my cluster has nodes in seven racks, and I have provided a rack topology script. What is Hadoop's block placement policy, assuming a block replication factor of three?

Answer 1

In your case, One copy of the block is written to a node in one rack; two copies are written to two nodes in a different rack.
Actually, what happens is HDFS uses rack-aware replica placement policy. In default configuration there are total 3 copies of a datablock on HDFS, 2 copies are stored on datanodes on same rack and 3rd copy on a different rack.

Hope this helps!