hadoop2.3.0单点伪分布与多点分布的配置

我的机器是mac book，安装virtualbox4.3.6，virtualbox安装ubunt13.10，在多点分布环境中，配置好一个机器后，clone出另外2个，一共三台机器。

1. Configure the Environment

Bash语言: 

sudo ./debian_set_java.sh jdk-7u51-linux-x64.tar.gz
sudo apt-get install -y openssh-server
sudo addgroup hadoop
sudo adduser —ingroup hadoop hadoop # create password
sudo visudo
hadoop  ALL=(ALL) ALL # hadoop user can use sudo
su - hadoop # need password
ssh-keygen -t rsa -P "" # Enter file (/home/hadoop/.ssh/id_rsa)

cat /home/hadoop/.ssh/id_rsa.pub >> /home/hadoop/.ssh/authorized_keys
wget
tar zxvf hadoop-2.3.0.tar.gz
sudo cp -pr hadoop-2.3.0/ /opt
cd /opt
sudo ln -s hadoop-2.3.0 hadoop
sudo chown -R hadoop:hadoop hadoop-2.3.0
sed -i 's/${JAVA_HOME}/\/usr\/local\/java\/jdk1.7.0_51/' hadoop/etc/hadoop/hadoop-env.sh

debian_set_java.sh

#!/bin/bash

if [ `whoami` != "root" ]; then
 echo "Use root to set java environment."
 exit 1
fi
if [ `uname` == 'Darwin' ]; then
 echo "Darwin system."
 exit 1
fi

function unpack_java()
{
 mkdir /usr/local/java
 tar -pzxf $1 -C /usr/local/java/
 if [ `echo $?` != 0 ]; then
   echo "untar file failed."
   exit 1
 fi  
 echo "unpack $1 done"
}

function set_j8_env()
{
 cp -p /etc/profile /tmp/profile.bak_`date "+%Y%m%d%H%M%S"`
 sed -i '$a \\nJAVA_HOME=/usr/local/java/jdk1.8.0\nPATH=$PATH:$JAVA_HOME/bin\nexport JAVA_HOME\nexport PATH'/etc/profile
 source /etc/profile
 update-alternatives --install /usr/bin/javac javac /usr/local/java/jdk1.8.0/bin/javac  95  
 update-alternatives --install /usr/bin/java  java  /usr/local/java/jdk1.8.0/bin/java  95  
 update-alternatives --config java
 echo "set java 8 environment done"
}

function set_j7_env()
{
 update-alternatives --install /usr/bin/javac javac /usr/local/java/jdk1.7.0_51/bin/javac  95  
 update-alternatives --install /usr/bin/java  java  /usr/local/java/jdk1.7.0_51/bin/java  95  
 update-alternatives --config java
 echo "set java 7 environment done"
}

unpack_java $1
set_j7_env

2. Configure hadoop single Node environment
cp mapred-site.xml.template mapred-site.xml
vi mapred-site.xml
 
    mapreduce.cluster.temp.dir
   
    No description
    true
 
   
    mapreduce.cluster.local.dir
   
    No description
    true
 


vi yarn-site.xml
 
    yarn.resourcemanager.resource-tracker.address
    127.0.0.1:8021
    host is the hostname of the resource manager and port is the port on which the NodeManagers contact the Resource Manager.
   
 
 
    yarn.resourcemanager.scheduler.address
    127.0.0.1:8022
    host is the hostname of the resourcemanager and port is the port on which the Applications in the cluster talk to the Resource Manager.
   
 
 
    yarn.resourcemanager.scheduler.class
    org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler
    In case you do not want to use the default scheduler
 
 
    yarn.resourcemanager.address
    127.0.0.1:8023
    the host is the hostname of the ResourceManager and the port is the port on which the clients can talk to the Resource Manager.
 
 
    yarn.nodemanager.local-dirs
   
    the local directories used by the nodemanager
 
 
    yarn.nodemanager.address
    0.0.0.0:8041
    the nodemanagers bind to this port
   
 
    yarn.nodemanager.resource.memory-mb
    10240
    the amount of memory on the NodeManager in GB
 
 
    yarn.nodemanager.remote-app-log-dir
    /app-logs
    directory on hdfs where the application logs are moved to
 
   
    yarn.nodemanager.log-dirs
   
    the directories used by Nodemanagers as log directories
 
 
    yarn.nodemanager.aux-services
    mapreduce_shuffle
    shuffle service that needs to be set for Map Reduce to run
 


补充配置：
mapred-site.xml
 
    mapreduce.framework.name
    yarn
 

core-site.xml
 
    fs.defaultFS
    hdfs://127.0.0.1:9000
 

hdfs-site.xml
 
    dfs.replication
    1
 

Bash语言: 

cd /opt/hadoop
bin/hdfs namenode -format
sbin/hadoop-daemon.sh start namenode
sbin/hadoop-daemon.sh start datanode
sbin/yarn-daemon.sh start resourcemanager
sbin/yarn-daemon.sh start nodemanager
jps

# Run a job on this node
bin/hadoop jar share/hadoop/mapreduce/hadoop-mapreduce-examples-2.3.0.jar pi 5 10

3. Running Problem
14/01/04 05:38:22 INFO ipc.Client: Retrying connect to server: localhost/127.0.0.1:8023. Already tried 9 time(s); retry policy is RetryUpToMaximumCountWithFixedSleep(maxRetries=10, sleepTime=1 SECONDS)


netstat -atnp # found tcp6
Solve:
cat /proc/sys/net/ipv6/conf/all/disable_ipv6 # 0 means ipv6 is on, 1 means off
cat /proc/sys/net/ipv6/conf/lo/disable_ipv6
cat /proc/sys/net/ipv6/conf/default/disable_ipv6
ip a | grep inet6 # have means ipv6 is on


sudo vi /etc/sysctl.conf
net.ipv6.conf.all.disable_ipv6=1  
net.ipv6.conf.default.disable_ipv6=1  
net.ipv6.conf.lo.disable_ipv6=1

sudo sysctl -p # have the same effect with reboot





4. Cluster setup

Config /opt/hadoop/etc/hadoop/{hadoop-env.sh, yarn-env.sh}

export JAVA_HOME=/usr/local/java/jdk1.7.0_51

cd /opt/hadoop

mkdir opt/hadoop/tmp

mkdir -p opt/hadoop/{data,name} # on every node. name on namenode, data on datanode

vi /etc/hosts # hostname also changed on each node

192.168.1.110  cloud1

192.168.1.112  cloud2

192.168.1.114  cloud3

vi /opt/hadoop/etc/hadoop/slaves

cloud2

cloud3

core-site.xml

 
    fs.defaultFS
    hdfs://cloud1:9000
 
 
    io.file.buffer.size
    131072
 
 
    hadoop.tmp.dir
    /opt/hadoop/tmp
    A base for other temporary directories.
 



据说dfs.datanode.data.dir 需要清空，不然datanode不能启动
hdfs-site.xml
 
    dfs.namenode.name.dir
    /opt/hadoop/name
 
 
    dfs.datanode.data.dir
    /opt/hadoop/data
 
 
    dfs.replication
    2
 


yarn-site.xml
 
    yarn.resourcemanager.address
    cloud1:8032
    ResourceManager host:port for clients to submit jobs.
 
 
    yarn.resourcemanager.scheduler.address
    cloud1:8030
    ResourceManager host:port for ApplicationMasters to talk to Scheduler to obtain resources.
 
 
    yarn.resourcemanager.resource-tracker.address
    cloud1:8031
    ResourceManager host:port for NodeManagers.
 
 
    yarn.resourcemanager.admin.address
    cloud1:8033
    ResourceManager host:port for administrative commands.
 
 
    yarn.resourcemanager.webapp.address
    cloud1:8088
    ResourceManager web-ui host:port.
 
 
    yarn.resourcemanager.scheduler.class
    org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler
    In case you do not want to use the default scheduler
 

 
    yarn.nodemanager.resource.memory-mb
    10240
    the amount of memory on the NodeManager in MB
 
 
    yarn.nodemanager.local-dirs
   
    the local directories used by the nodemanager
 
 
    yarn.nodemanager.log-dirs
   
    the directories used by Nodemanagers as log directories
 
 
    yarn.nodemanager.remote-app-log-dir
    /app-logs
    directory on hdfs where the application logs are moved to
 
 
    yarn.nodemanager.aux-services
    mapreduce_shuffle
    shuffle service that needs to be set for Map Reduce to run
 



mapred-site.xml
 
    mapreduce.framework.name
    yarn
 
 
    mapreduce.jobhistory.address
    cloud1:10020
 
 
    mapreduce.jobhistory.webapp.address
    cloud1:19888
 


cd /opt/hadoop/
bin/hdfs namenode -format
sbin/start-dfs.sh # cloud1 NameNode SecondaryNameNode, cloud2 and cloud3 DataNode
sbin/start-yarn.sh # cloud1 ResourceManager, cloud2 and cloud3 NodeManager
jps

查看集群状态 bin/hdfs dfsadmin -report
查看文件块组成 bin/hdfs fsck / -files -blocks
NameNode查看hdfs  
查看RM    
bin/hdfs dfs -mkdir /input
bin/hadoop jar ./share/hadoop/mapreduce/hadoop-mapreduce-examples-2.3.0.jar randomwriter input


5. Questions:
14/01/05 23:59:05 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
/opt/hadoop/lib/native/ 下面的动态链接库是32bit的，要替换成64位的

ssh 登录出现Are you sure you want to continue connecting (yes/no)?解决方法
修改/etc/ssh/ssh_config 将其中的# StrictHostKeyChecking ask 改成 StrictHostKeyChecking no

两个slaves的DataNode无法加入cluster系统，把/etc/hosts 里面127.0.1.1或localhost 的内容行删除