Hadoop v2 YARN deployment includes deploying the ResourceManager service on the master node and deploying NodeManager services in the slave nodes. YARN ResourceManager is the service that arbitrates all the resources of the cluster, and NodeManager is the service that manages the resources in a single node.
Hadoop MapReduce applications can run on YARN using a YARN ApplicationMaster to coordinate each job and a set of resource containers to run the Map and Reduce tasks.
Tip
Installing Hadoop directly using Hadoop release artifacts, as mentioned in this recipe, is recommended for development testing and for advanced use cases only. For regular production clusters, we recommend using a packaged Hadoop distribution as mentioned in the Setting up Hadoop ecosystem in a distributed cluster environment using a Hadoop distribution recipe. Packaged Hadoop distributions make it much easier to install, configure, maintain, and update the components of the Hadoop ecosystem.
You can follow this recipe either using a single machine as a pseudo-distributed installation or using a multiple machine cluster. If you are using multiple machines, you should choose one machine as the master node where you will run the HDFS NameNode and the YARN ResourceManager. If you are using a single machine, use it as both the master node as well as the slave node.
Set up HDFS by following the Setting up HDFS recipe.
Let's set up Hadoop YARN by setting up the YARN ResourceManager and the NodeManagers.
In each machine, create a directory named local inside
{HADOOP_DATA_DIR}, whichyou created in the Setting up HDFS recipe. Change the directory permissions to755.Add the following to the
{HADOOP_HOME}/etc/hadoop/mapred-site.xmltemplate and save it as{HADOOP_HOME}/etc/hadoop/mapred-site.xml:<property> <name>fs.default.name</name> <value>hdfs://localhost:9000</value> </property>
Add the following to the
{HADOOP_HOME}/etc/hadoop/yarn-site.xmlfile:<property> <name>yarn.nodemanager.aux-services</name> <value>mapreduce_shuffle</value> </property> <property> <name>yarn.nodemanager.aux-services.mapreduce_shuffle.class</name> <value>org.apache.hadoop.mapred.ShuffleHandler</value> </property>
Start HDFS using the following command:
$ $HADOOP_HOME/sbin/start-dfs.shRun the following command to start the YARN services:
$ $HADOOP_HOME/sbin/start-yarn.sh starting yarn daemons starting resourcemanager, logging to ……… xxx.xx.xxx.xxx: starting nodemanager, logging to ………
Run the following command to start the MapReduce JobHistoryServer. This enables the web console for MapReduce job histories:
$ $HADOOP_HOME/sbin/mr-jobhistory-daemon.sh start historyserverVerify the installation by listing the processes through the
jpscommand. The master node will list the NameNode, ResourceManager, and JobHistoryServer services. The slave nodes will list DataNode and NodeManager services:$ jps 27084 NameNode 2073 JobHistoryServer 2106 Jps 2588 1536 ResourceManager
Visit the web-based monitoring pages for ResourceManager available at
http://{MASTER_NODE}:8088/.
As described in the introduction to the chapter, Hadoop v2 installation consists of HDFS nodes, YARN ResourceManager, and worker nodes. When we start the NameNode, it finds slaves through the HADOOP_HOME/slaves file and uses SSH to start the DataNodes in the remote server at the startup. Also, when we start ResourceManager, it finds slaves through the HADOOP_HOME/slaves file and starts NodeManagers.