Define DataNode. How does NameNode tackle DataNode failures?

DataNode is a node in HDFS that holds data; it is a node where real data is stored in the file system. To indicate that it is alive, each DataNode sends a heartbeat message. If the NameNode does not receive a message from the DataNode after 10 minutes, it deems the DataNode to be dead or out of place and begins replication of blocks that were previously hosted on that DataNode to another DataNode. A BlockReport is a report that lists all of the blocks on a DataNode. The system now begins to duplicate the contents of the DataNode that has died.

The NameNode is in charge of replicating data blocks from one DataNode to the next. The replication data is sent directly between DataNodes in this procedure, and it never travels via the NameNode.