Steps to install Hadoop 2.x release (Yarn or Next-Gen) on multi-node cluster

In the previous post, we saw how to setup Hadoop 2.x on single-node. Here, we will see how to set up a multi-node cluster.

Hadoop 2.x release involves many changes to Hadoop and MapReduce. The centralized JobTracker service is replaced with a ResourceManager that manages the resources in the cluster and an ApplicationManager that manages the application lifecycle. These architectural changes enable hadoop to scale to much larger clusters. For more details on architectural changes in Hadoop next-gen (a.k.a. Yarn), watch this video or visit this blog.

This post concentrates on installing Hadoop 2.x a.k.a. Yarn a.k.a. next-gen on a multi-node cluster.

Prerequisites:

  • Java 6 installed
  • Dedicated user for hadoop
  • SSH configured

Steps to install Hadoop 2.x:

1. Download tarball

You can download tarball for hadoop 2.x from here. Extract it to a folder say, /home/hduser/yarn on master and all the slaves. We assume dedicated user for Hadoop is “hduser”.

NOTE: Master and all the slaves must have the same user and hadoop directory on same path.

$ cd /home/hduser/yarn
$ sudo chown -R hduser:hadoop hadoop-2.0.1-alpha

2. Edit /etc/hosts

Add the association between the hostnames and the ip address for the master and the slaves on all the nodes in the /etc/hosts file. Make sure that the all the nodes in the cluster are able to ping to each other.

Important Change:

127.0.0.1 localhost localhost.localdomain my-laptop
127.0.1.1 my-laptop

If you have provided alias for localhost (as done in entries above), protocol buffers will try to connect to my-laptop from other hosts while making RPC calls which will fail.

Solution:

Assuming the machine (my-laptop) has ip address “10.3.3.43”, make an entry as follows in all the other machines:

10.3.3.43       my-laptop

3. Password less SSH

Make sure that the master is able to do a password-less ssh to all the slaves.

4. Edit ~/.bashrc

export HADOOP_HOME=/home/hduser/yarn/hadoop-2.0.1-alpha
export HADOOP_MAPRED_HOME=$HADOOP_HOME
export HADOOP_COMMON_HOME=$HADOOP_HOME
export HADOOP_HDFS_HOME=$HADOOP_HOME
export YARN_HOME=$HADOOP_HOME
export HADOOP_CONF_DIR=$HADOOP_HOME/etc/hadoop
export YARN_CONF_DIR=$HADOOP_HOME/etc/hadoop

5. Edit Hadoop environment files

Add JAVA_HOME to following files

Add following line at start of script in libexec/hadoop-config.sh :

export JAVA_HOME=/usr/lib/jvm/java-6-openjdk-i386/

Add following lines at start of script in etc/hadoop/yarn-env.sh :

export JAVA_HOME=/usr/lib/jvm/java-6-openjdk-i386/
export HADOOP_HOME=/home/hduser/yarn/hadoop-2.0.1-alpha
export HADOOP_MAPRED_HOME=$HADOOP_HOME
export HADOOP_COMMON_HOME=$HADOOP_HOME
export HADOOP_HDFS_HOME=$HADOOP_HOME
export YARN_HOME=$HADOOP_HOME
export HADOOP_CONF_DIR=$HADOOP_HOME/etc/hadoop
export YARN_CONF_DIR=$HADOOP_HOME/etc/hadoop

Change the path as per your java installation.

6. Create Temp folder in HADOOP_HOME

$ mkdir -p $HADOOP_HOME/tmp

7. Add properties in configuration files

Make changes as mentioned below in all the machines:

$HADOOP_CONF_DIR/core-site.xml

<?xml version="1.0" encoding="UTF-8"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<configuration>
  <property>
    <name>fs.default.name</name>
    <value>hdfs://master:9000</value>
  </property>
  <property>
    <name>hadoop.tmp.dir</name>
    <value>/home/hduser/yarn/hadoop-2.0.1-alpha/tmp</value>
  </property>
</configuration>

$HADOOP_CONF_DIR/hdfs-site.xml :

<?xml version="1.0" encoding="UTF-8"?>
 <?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
 <configuration>
   <property>
     <name>dfs.replication</name>
     <value>2</value>
   </property>
   <property>
     <name>dfs.permissions</name>
     <value>false</value>
   </property>
 </configuration>

$HADOOP_CONF_DIR/mapred-site.xml :

<?xml version="1.0"?>
<configuration>
 <property>
   <name>mapreduce.framework.name</name>
   <value>yarn</value>
 </property>
</configuration>

$HADOOP_CONF_DIR/yarn-site.xml :

<?xml version="1.0"?>
 <configuration>
  <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>
  <property>
    <name>yarn.resourcemanager.resource-tracker.address</name>
    <value>master:8025</value>
  </property>
  <property>
    <name>yarn.resourcemanager.scheduler.address</name>
    <value>master:8030</value>
  </property>
  <property>
    <name>yarn.resourcemanager.address</name>
    <value>master:8040</value>
  </property>
 </configuration>

8. Add slaves

Add the slave entries in $HADOOP_CONF_DIR/slaves on master machine:

slave1
slave2

9. Format the namenode

$ bin/hadoop namenode -format

10. Start Hadoop Daemons

$ sbin/hadoop-daemon.sh start namenode
$ sbin/hadoop-daemons.sh start datanode
$ sbin/yarn-daemon.sh start resourcemanager
$ sbin/yarn-daemons.sh start nodemanager
$ sbin/mr-jobhistory-daemon.sh start historyserver

NOTE: For datanode and nodemanager, scripts are *-daemons.sh and not *-daemon.sh. daemon.sh does not lookup in slaves file and hence, will only start processes on master

11. Check installation

Check for jps output on slaves and master.

For master:

$ jps
6539 ResourceManager
6451 DataNode
8701 Jps
6895 JobHistoryServer
6234 NameNode
6765 NodeManager

For slaves:

$ jps
8014 NodeManager
7858 DataNode
9868 Jps

If these services are not up, check the logs in $HADOOP_HOME/logs directory to identify the issue.

12. Run a demo application to verify installtion

$ mkdir in
$ cat > in/file
This is one line
This is another one

Add this directory to HDFS:

$ bin/hadoop dfs -copyFromLocal in /in

Run wordcount example provided:

$ bin/hadoop jar share/hadoop/mapreduce/hadoop-mapreduce-examples-2.*-alpha.jar wordcount /in /out

Check the output:

$ bin/hadoop dfs -cat /out/*
This 2
another 1
is 2
line 1
one 2

13. Web interface

1. http://master:50070/dfshealth.jsp
2. http://master:8088/cluster
3. http://master:19888/jobhistory (for Job History Server)

14. Stopping the daemons

$ sbin/mr-jobhistory-daemon.sh stop historyserver
$ sbin/yarn-daemons.sh stop nodemanager
$ sbin/yarn-daemon.sh stop resourcemanager
$ sbin/hadoop-daemons.sh stop datanode
$ sbin/hadoop-daemon.sh stop namenode

15. Possible errors

If you get a exception stack trace similar to given below:

Container launch failed for container_1350204169962_0002_01_000004 : java.lang.reflect.UndeclaredThrowableException
 at org.apache.hadoop.yarn.exceptions.impl.pb.YarnRemoteExceptionPBImpl.unwrapAndThrowException(YarnRemoteExceptionPBImpl.java:135)
 at org.apache.hadoop.yarn.api.impl.pb.client.ContainerManagerPBClientImpl.startContainer(ContainerManagerPBClientImpl.java:101)
 at org.apache.hadoop.mapreduce.v2.app.launcher.ContainerLauncherImpl$Container.launch(ContainerLauncherImpl.java:149)
 at org.apache.hadoop.mapreduce.v2.app.launcher.ContainerLauncherImpl$EventProcessor.run(ContainerLauncherImpl.java:373)
 at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1110)
 at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:603)
 at java.lang.Thread.run(Thread.java:679)
Caused by: com.google.protobuf.ServiceException: java.net.UnknownHostException: Invalid host name: local host is: (unknown); destination host is: "my-laptop":40365; java.net.UnknownHostException; For more details see: http://wiki.apache.org/hadoop/UnknownHost
 at org.apache.hadoop.ipc.ProtobufRpcEngine$Invoker.invoke(ProtobufRpcEngine.java:187)
 at $Proxy29.startContainer(Unknown Source)
 at org.apache.hadoop.yarn.api.impl.pb.client.ContainerManagerPBClientImpl.startContainer(ContainerManagerPBClientImpl.java:99)
 ... 5 more
Caused by: java.net.UnknownHostException: Invalid host name: local host is: (unknown); destination host is: "my-laptop":40365; java.net.UnknownHostException; For more details see: http://wiki.apache.org/hadoop/UnknownHost
 at org.apache.hadoop.net.NetUtils.wrapException(NetUtils.java:740)
 at org.apache.hadoop.ipc.Client$Connection.<init>(Client.java:248)
 at org.apache.hadoop.ipc.Client.getConnection(Client.java:1261)
 at org.apache.hadoop.ipc.Client.call(Client.java:1141)
 at org.apache.hadoop.ipc.ProtobufRpcEngine$Invoker.invoke(ProtobufRpcEngine.java:184)
 ... 7 more
Caused by: java.net.UnknownHostException
 ... 11 more

Solution: Check the Important Change in Step 2 and apply the necessary changes.

 

Happy Coding!!!

– Rasesh Mori

Advertisements

Steps to install Hadoop 2.x release (Yarn or Next-Gen) on single node cluster setup

Hadoop 2.x release involves many changes to Hadoop and MapReduce. The centralized JobTracker service is replaced with a ResourceManager that manages the resources in the cluster and an ApplicationManager that manages the application lifecycle. These architectural changes enable hadoop to scale to much larger clusters. For more details on architectural changes in Hadoop next-gen (a.k.a. Yarn), watch this video or visit this blog.

This post concentrates on installing Hadoop 2.x a.k.a. Yarn a.k.a. next-gen on a single-node cluster.

Prerequisites:

  • Java 6 installed
  • Dedicated user for hadoop
  • SSH configured

Steps to install Hadoop 2.x:

1. Download tarball

You can download tarball for hadoop 2.x from here. Extract it to a folder say, /home/hduser/yarn. We assume dedicated user for Hadoop is “hduser”.

$ cd /home/hduser/yarn
$ sudo chown -R hduser:hadoop hadoop-2.0.1-alpha

2. Setup Environment Variables

$ export HADOOP_HOME=$HOME/yarn/hadoop-2.0.1-alpha
$ export HADOOP_MAPRED_HOME=$HOME/yarn/hadoop-2.0.1-alpha
$ export HADOOP_COMMON_HOME=$HOME/yarn/hadoop-2.0.1-alpha
$ export HADOOP_HDFS_HOME=$HOME/yarn/hadoop-2.0.1-alpha
$ export HADOOP_YARN_HOME=$HOME/yarn/hadoop-2.0.1-alpha
$ export HADOOP_CONF_DIR=$HOME/yarn/hadoop-2.0.1-alpha/etc/hadoop

This is very important as if you miss any one variable or set the value incorrectly, it will be very difficult to detect the error and the job will fail.

Also, add these to your ~/.bashrc or other shell start-up script so that you don’t need to set them every time.

3. Create directories

Create two directories to be used by namenode and datanode.

$ mkdir -p $HOME/yarn/yarn_data/hdfs/namenode
$ mkdir -p $HOME/yarn/yarn_data/hdfs/datanode

4. Set up config files

$ cd $YARN_HOME

Add the following properties under configuration tag in the files mentioned below:

etc/hadoop/yarn-site.xml:

   yarn.nodemanager.aux-services
   mapreduce_shuffle


   yarn.nodemanager.aux-services.mapreduce.shuffle.class
   org.apache.hadoop.mapred.ShuffleHandler

etc/hadoop/core-site.xml:

   fs.default.name
   hdfs://localhost:9000

etc/hadoop/hdfs-site.xml:

 
   dfs.replication
   1
 
 
   dfs.namenode.name.dir
   file:/home/hduser/yarn/yarn_data/hdfs/namenode
 
 
   dfs.datanode.data.dir
   file:/home/hduser/yarn/yarn_data/hdfs/datanode

etc/hadoop/mapred-site.xml:

If this file does not exist, create it and paste the content provided below:



   
      mapreduce.framework.name
      yarn
   

5. Format namenode

This step is needed only for the first time. Doing it every time will result in loss of content on HDFS.

$ bin/hadoop namenode -format

6. Start HDFS processes

Name node:

$ sbin/hadoop-daemon.sh start namenode
starting namenode, logging to /home/hduser/yarn/hadoop-2.0.1-alpha/logs/hadoop-hduser-namenode-pc3-laptop.out
$ jps
18509 Jps
17107 NameNode

Data node:

$ sbin/hadoop-daemon.sh start datanode
starting datanode, logging to /home/hduser/yarn/hadoop-2.0.1-alpha/logs/hadoop-hduser-datanode-pc3-laptop.out
$ jps
18509 Jps
17107 NameNode
17170 DataNode

7. Start Hadoop Map-Reduce Processes

Resource Manager:

$ sbin/yarn-daemon.sh start resourcemanager
starting resourcemanager, logging to /home/hduser/yarn/hadoop-2.0.1-alpha/logs/yarn-hduser-resourcemanager-pc3-laptop.out
$ jps
18509 Jps
17107 NameNode
17170 DataNode
17252 ResourceManager

Node Manager:

$ sbin/yarn-daemon.sh start nodemanager
starting nodemanager, logging to /home/hduser/yarn/hadoop-2.0.1-alpha/logs/yarn-hduser-nodemanager-pc3-laptop.out
$jps
18509 Jps
17107 NameNode
17170 DataNode
17252 ResourceManager
17309 NodeManager

Job History Server:

$ sbin/mr-jobhistory-daemon.sh start historyserver
starting historyserver, logging to /home/hduser/yarn/hadoop-2.0.1-alpha/logs/yarn-hduser-historyserver-pc3-laptop.out
$jps
18509 Jps
17107 NameNode
17170 DataNode
17252 ResourceManager
17309 NodeManager
17626 JobHistoryServer

8. Running the famous wordcount example to verify installation

$ mkdir in
$ cat > in/file
This is one line
This is another one

Add this directory to HDFS:

$ bin/hadoop dfs -copyFromLocal in /in

Run wordcount example provided:

$ bin/hadoop jar share/hadoop/mapreduce/hadoop-mapreduce-examples-2.*-alpha.jar wordcount /in /out

Check the output:

$ bin/hadoop dfs -cat /out/*
This 2
another 1
is 2
line 1
one 2

9. Web interface

Browse HDFS and check health using http://localhost:50070 in the browser:

You can check the status of the applications running using the following URL:

http://localhost:8088

10. Stop the processes

$ sbin/hadoop-daemon.sh stop namenode
$ sbin/hadoop-daemon.sh stop datanode
$ sbin/yarn-daemon.sh stop resourcemanager
$ sbin/yarn-daemon.sh stop nodemanager
$ sbin/mr-jobhistory-daemon.sh stop historyserver

Happy Coding!!!