\\
**[[cluster:0|Back]]**

==== Build Hadoop (test) Cluster ====

[[cluster:115|Use Hadoop (test) Cluster]]

These are my notes building a test Hadoop cluster on virtual machines in VMware. They consists of a blending of instructions posted by others with my commentary added.  Please review these sites so this page makes sense to you.

  * CTOvision [[http://ctovision.com/2012/01/cloudera-hadoop-quickstart/]] 
  * Dakini [[http://dak1n1.com/blog/9-hadoop-el6-install]]

Other sites you want to read:

  * Yahoo [[http://developer.yahoo.com/hadoop/tutorial/]]
  * Apache [[http://hadoop.apache.org/docs/r1.0.4/index.html]]
  * Noll [[http://www.michael-noll.com/tutorials/running-hadoop-on-ubuntu-linux-multi-node-cluster/]]
  * IBM article [[http://www.ibm.com/developerworks/data/library/techarticle/dm-1209hadoopbigdata/]]

And

  * White [[http://hadoopbook.com/]]

==== Building ====

  * Deployed 8 virtual machines, Oracle Linux 6, 64 bit, bare bones.
  * Each node has 1 GB ram and a 36 GB hard disk.

  * First get rid of OpenJDK if it's in your VMware template
    * Consult CTOvision on how to do that.
  * The download latest Java packages from Oracle and install.
  * Everything below is done by root.

<code>
# all nodes, i used pdsh to spawn commands across all nodes
rpm -ivh /usr/local/src/jdk-7u21-linux-x64.rpm                    
rpm -ivh /usr/local/src/jre-7u21-linux-x64.rpm                                                                        
alternatives --install /usr/bin/java java /usr/java/latest/bin/java 1600 
alternatives --auto java
# fix this as some Hadoop scripts look at this location
cd /usr/java
ln -s ./latest/bin
which java 
java -version
</code>                                              

  * Next set up the Cloudera repository

<code>
# all nodes
cd /etc/yum.repos.d/
wget http://archive.cloudera.com/redhat/6/x86_64/cdh/cloudera-cdh3.repo
yum update                                                             
yum install hadoop-0.20
</code>                                              

  * Selinux, again ...

<code>
setenforce 0
# edit this file and disable  
vi /etc/selinux/config 
</code>

  * Ports, the node need to talk to each other as well as allow admin pages to load

<code>
# edit this file and restart iptables     
vi /etc/sysconfig/iptables
# hadoop
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 50070 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 50075 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 50090 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 50105 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 50030 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 50060 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 8020 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 50010 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 50020 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 50100 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 8021 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 9001 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 8012 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 54310 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -m iprange --src-range 129.133.x.xxx-129.133.x.xxx --dport 54311 -j ACCEPT
# plus 127.0.0.1:0 and maybe 9000
# hadoop admin status
-A INPUT -m state --state NEW -m tcp -p tcp -s 129.133.0.0/16 --dport 50030 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -s 129.133.0.0/16 --dport 50070 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp -s 129.133.0.0/16 --dport 50075 -j ACCEPT
</code>

  * Install the **headnode** node and tracker

<code>
# head node
yum -y install hadoop-0.20-namenode
yum -y install hadoop-0.20-jobtracker
</code>

  * On all the work nodes

<code>
# data node
yum -y install hadoop-0.20-datanode                                    
yum -y install hadoop-0.20-tasktracker   
</code>                              
     
  * Next set up the configuration environment
  * Edit the conf files, consult Dakini site for content
  * Copy those 3 files to all work nodes
  * The display command should point to the MyCluster files

<code>
# all nodes                                                            
cp -r /etc/hadoop-0.20/conf.empty   /etc/hadoop-0.20/conf.MyCluster
alternatives --install /etc/hadoop-0.20/conf hadoop-0.20-conf /etc/hadoop-0.20/conf.MyCluster 50
alternatives --set hadoop-0.20-conf /etc/hadoop-0.20/conf.MyCluster
alternatives --display hadoop-0.20-conf
vi /etc/hadoop-0.20/conf.MyCluster/core-site.xml
vi /etc/hadoop-0.20/conf.MyCluster/hdfs-site.xml
vi /etc/hadoop-0.20/conf.MyCluster/mapred-site.xml
</code>

  * Since this is a test cluster I located the DHFS filesystem on the OS disk
      * In a production environment you'd want multiple dedicated disks per node

<code>
# all nodes
mkdir -p /mnt/hdfs/1
mkdir -p /mnt/hdfs/1/namenode
mkdir -p /mnt/hdfs/1/datanode
mkdir -p /mnt/hdfs/1/mapred
chown -R hdfs:hadoop /mnt/hdfs
chown -R mapred:hadoop /mnt/hdfs/1/mapred
</code>

  * Format HDFS! Very important. Do ONLY ONCE on head node.

<code>
# headnode only
sudo -u hdfs hadoop namenode -format
</code>

  * Fix permissions

<code>
# all nodes
chgrp hdfs /usr/lib/hadoop-0.20/
chmod g+rw /usr/lib/hadoop-0.20/
</code>

  * Start Hadoop nodes and trackers
  * If you receive the dreaded "datanode dead but pid exists" error
  * Check the log in question it'll give a hint
    * You may have typos in the XML files and configuration does not load
    * File permissions may prevent nodes and trackers from starting
    * You missed a step, like in the alternatives commands
    * You issued the HDFS format command multiples times

<code>
# head node
/etc/init.d/hadoop-0.20-namenode start
/etc/init.d/hadoop-0.20-jobtracker start

# work nodes
/etc/init.d/hadoop-0.20-datanode start
/etc/init.d/hadoop-0.20-tasktracker start
</code>

  * Alright, lets some filesystem entries

<code>
# head node only
sudo -u hdfs hadoop fs -mkdir /mapred/system
sudo -u hdfs hadoop fs -chown mapred:hadoop /mapred/system
sudo -u hdfs hadoop dfs -mkdir /tmp
sudo -u hdfs hadoop dfs -chmod -R 1777 /tmp
</code>

  * Command line health check

<code>
sudo -u hdfs hadoop dfsadmin -report
sudo -u hdfs hadoop dfs -df
</code>

  * And from a remote machine access your head node
    * Hadoop Map/Reduce Administration
      * [[http://headnode.wesleyan.edu:50030]] 
    * The Namenode
      * [[http://headnode.wesleyan.edu:50070]]

TODO

  * Run some jobs
  * Find a MOOC course

\\
**[[cluster:0|Back]]**