Cloud

Manual Cluster setup

Back to Hadoop section

We are going to install Hadoop 2.0 (YARN + MR2) on Virtual Machines (VMs) of the StratusLab Cloud. We will use 2 VMs (master + slave) where the master will be also a slave.

Launch VMs

$ stratus-run-instance --cpu=4 --ram=16384 --swap=2048 --vm-name=master 
--volatile-disk=20 ID
$ stratus-run-instance --cpu=4 --ram=16384 --swap=2048 --vm-name=slave 
--volatile-disk=20 ID

Connect and config each VM

Mount external disk

$ mkdir /mnt/data_hadoop
$ mount /dev/vdc /mnt/data_hadoop/

Use automatic configuration

$ source custom.sh

Ckeck if iptable is not running

$ service iptables status

Copy ssh keys of each MV

MV1$ cat /home/hadoop/.ssh/id_rsa.pub 
MV2$ nano /home/hadoop/.ssh/authorized_keys

Configure host file

 $ more /etc/hosts
134.158.75.XX onevm-XX.lal.in2p3.fr master
134.158.75.XX onevm-XX.lal.in2p3.fr slave

Configure Hadoop files

  • Parameters file:
core-site.xml   hdfs-site.xml   mapred-site.xml yarn-site.xml
  • Only on Master node:
$ cat $HADOOP_INSTALL/etc/hadoop/masters
master
$ cat $HADOOP_INSTALL/etc/hadoop/slaves 
master
slave

Start Hadoop services

Format HDFS from master

$ hdfs namenode -format
...
Storage directory /mnt/data_hadoop/hadoop/tmp/dfs/name has been successfully formatted

Start Hadoop daemons

  • Only on Master node:
$HADOOP_INSTALL/sbin/hadoop-daemon.sh start namenode
$HADOOP_INSTALL/sbin/hadoop-daemon.sh start datanode 
$HADOOP_INSTALL/sbin/yarn-daemon.sh start resourcemanager
$HADOOP_INSTALL/sbin/yarn-daemon.sh start nodemanager 
$HADOOP_INSTALL/sbin/yarn-daemon.sh start proxyserver
$HADOOP_INSTALL/sbin/mr-jobhistory-daemon.sh start historyserver  
  • Only on Slave node:
$HADOOP_INSTALL/sbin/hadoop-daemon.sh start datanode 
$HADOOP_INSTALL/sbin/yarn-daemon.sh start nodemanager

Check

Master$ jps
NameNode
DataNode
ResourceManager
NodeManager
JobHistoryServer

Slave$ jps
DataNode
NodeManager

API

http://master:50070 =>  Namenode 
http://master:8088 => ResourceManager (cluster)
http://master:50090 => Secondary NameNode
http://master:50075 => DataNode