Question

2 datanodes is slower ,how to get those details .what is your approach as hadoop admin ?

Answer 1

First, enable the detection of slow nodes. You can do this by setting the value of dfs.namenode.peer.stats.enabled property to true. 

<property>
 <name>dfs.datanode.peer.stats.enabled</name>
 <value>true</value> 
</property>

Let the cluster run for sometime. 

To access the details, visit: http://<namenode_host>:50070/jmx

Here you will see the time taken by datanodes. After you have detected the datanodes that are slow, you have to check for the bottlenecks in data flow.