For 1#, since you still have 2 datanodes alive, and the replication is 2, writing will success. (Read will success) For 2#, now you only have 1 datanode, and the replication is 2, then initial writing will success, but later sometime pipeline recovery will fail.
Regards, Yi Liu -----Original Message----- From: Satyam Singh [mailto:satyam.si...@ericsson.com] Sent: Thursday, August 28, 2014 3:01 PM To: user@hadoop.apache.org Subject: replication factor in hdfs-site.xml Hi Users, I want behaviour for hadoop cluster for writing/reading in following replication cases: 1. replication=2 and in cluster (3 datatnodes+namenode) one datanode gets down. 2. replication=2 and in cluster (3 datatnodes+namenode) 2 datanode gets down. BR, Satyam