Client connects to the NameNode and ask about a file using the name of the file. NameNode looks up the block locations for the file and returns the same to the client. The client can then connect to the DataNodes and read the blocks needed. NameNode does not participate in the data transfer.

The following is the flow of a read request from a client. First, the client gets the locations and then pulls the blocks from the DataNodes. If a DataNode fails in the middle, then the client gets the replica of the block from another DataNode.