Hi Renjith, As per your setup description I assume you have one master node (namenode) and 2 datanodes. Please find answers below:-
1) If one datanode gets stopped or suspended, the heartbeat between the datanode and master gets stopped and then namenode(master) takes care of finding the next appropriate available datanode to replicate the lost blocks and will automatically trigger replication and hence maintains the replication factor. For more details you can check the below link:- https://hadoop.apache.org/docs/current/hadoop-project-dist/hadoop-hdfs/HdfsDesign.html#Data_Disk_Failure_Heartbeats_and_Re-Replication 2) If you have properly configured the master in master.txt and slaves in slaves.txt inside the <hadoop-dir>/etc/hadoop/ directory and if you have made sure ssh connectivity works fine between master and datanodes then all you have to do is run start-all.sh in the master node and the framework takes care of starting the datanode instances in the datanode servers on its own . Thanks, Manjunath On Tue, Jan 3, 2017 at 9:46 AM, Renjith Gk <[email protected]> wrote: > Hi All, > > As part of my lab exercise i am doing a self study of Hadoop. i have > cloned Master to 2 Datanodes as Datanode 1 and Datanode 2. > > Two Querys from Real Scenarios : > > 1. If Master is Up and running, DataNodes are stopped/suspended, will > there be any communication channel from Datanode to Master or will the > replication works ? > > 2.. Is there a need to run this command start-dfs.sh or start-all.sh in > master as well as datanodes ? > > Thanks, > Renjith >
