1、集群部署概述

1.1 Hadoop简介

    研发要做数据挖掘统计，需要Hadoop环境，便开始了本次安装测试，仅仅使用了3台虚拟机做测试工作。 简介……此处省略好多……，可自行查找 ……

    从你找到的内容可以总结看到，NameNode和JobTracker负责分派任务，DataNode和TaskTracker负责数据计算和存储。这样集群中可以有一台NameNode+JobTracker，N多台DataNode和TaskTracker。

1.2版本信息

本次测试安装所需软件版本信息如表1-1所示。

表1-1：软件版本信息

1.3测试环境说明

    本实验环境是在虚拟机中安装测试的，Hadoop集群中包括1个Master，2个Salve，节点之间内网互通，虚拟机主机名和IP地址如表1-2所示。

### 说明：文档出现的灰色阴影部分内容为文件编辑内容或操作显示内容。

2、操作系统设置

1、安装常用软件

### 由于操作系统是最小化安装，所以安装一些常用的软件包

# yum install gcc gcc-c++ openssh-clients vimmake ntpdate unzip cmake tcpdump openssl openssl-devel lzo lzo-devel zlibzlib-devel snappy snappy-devel lz4 lz4-devel bzip2 bzip2-devel cmake wget

2、修改主机名

# vim /etc/sysconfig/network        # 其他两个节点分别是：slave1和slave2

NETWORKING=yes

HOSTNAME=master

3、配置hosts文件

# vim /etc/hosts      # master和slave服务器上均添加以下配置内容

10.0.24.15 master

10.0.24.16 slave1

10.0.24.17 slave2

4、创建账号

# useradd hadoop

5、文件句柄设置

# vim/etc/security/limits.conf

*  soft nofile 65000

*  hard nofile 65535

$ ulimit -n   # 查看

6、系统内核参数调优sysctl.conf

net.ipv4.ip_forward = 0

net.ipv4.conf.default.rp_filter = 1

net.ipv4.conf.default.accept_source_route = 0

kernel.sysrq = 0

kernel.core_uses_pid = 1

net.ipv4.tcp_syncookies = 1

kernel.msgmnb = 65536

kernel.msgmax = 65536

kernel.shmmax = 68719476736

kernel.shmall = 4294967296

net.ipv4.tcp_max_tw_buckets = 60000

net.ipv4.tcp_sack = 1

net.ipv4.tcp_window_scaling = 1

net.ipv4.tcp_rmem = 4096 87380 4194304

net.ipv4.tcp_wmem = 4096 16384 4194304

net.core.wmem_default = 8388608

net.core.rmem_default = 8388608

net.core.rmem_max = 16777216

net.core.wmem_max = 16777216

net.core.netdev_max_backlog = 262144

net.core.somaxconn = 262144

net.ipv4.tcp_max_orphans = 3276800

net.ipv4.tcp_max_syn_backlog = 262144

net.ipv4.tcp_timestamps = 0

net.ipv4.tcp_synack_retries = 1

net.ipv4.tcp_syn_retries = 1

net.ipv4.tcp_tw_recycle = 1

net.ipv4.tcp_tw_reuse = 1

net.ipv4.tcp_mem = 94500000 915000000 927000000

net.ipv4.tcp_fin_timeout = 1

net.ipv4.tcp_keepalive_time = 1200

net.ipv4.tcp_max_syn_backlog = 65536

net.ipv4.tcp_timestamps = 0

net.ipv4.tcp_synack_retries = 2

net.ipv4.tcp_syn_retries = 2

net.ipv4.tcp_tw_recycle = 1

#net.ipv4.tcp_tw_len = 1

net.ipv4.tcp_tw_reuse = 1

#net.ipv4.tcp_fin_timeout = 30

#net.ipv4.tcp_keepalive_time = 120

net.ipv4.ip_local_port_range = 1024  65535

7、关闭SELINUX

# vim /etc/selinux/config  

#SELINUX=enforcing      

#SELINUXTYPE=targeted  

SELINUX=disabled       

# reboot                    # 重启服务器生效

8、配置ssh

# vim /etc/ssh/sshd_config       # 去掉以下内容前“#”注释

HostKey /etc/ssh/ssh_host_rsa_key

RSAAuthentication yes

PubkeyAuthentication yes

AuthorizedKeysFile      .ssh/authorized_keys

# /etc/init.d/sshd restart

9、配置master和slave间无密码互相登录

（1）maseter和slave服务器上均生成密钥

# su - hadoop

$ssh-keygen -b 1024 -t rsa

Generating public/private rsa key pair.

Enter file in which to save the key(/root/.ssh/id_rsa): <–直接输入回车

Enter passphrase (empty for no passphrase): <–直接输入回车

Enter same passphrase again: <–直接输入回车

Your identification has been saved in/root/.ssh/id_rsa.

Your public key has been saved in/root/.ssh/id_rsa.pub.

The key fingerprint is: ……

注意：在程序提示输入 passphrase 时直接输入回车，表示无证书密码。

（2）maseter和slave服务器上hadoop用户下均创建authorized_keys文件

$ cd .ssh

$ vim authorized_keys   # 添加master和salve服务器上hadoop用户下id_rsa.pub文件内容

ssh-rsa AAAAB3Nza…省略…HxNDk= hadoop@master

ssh-rsa AAAAB3Nza…省略…7CmlRs= hadoop@slave1

ssh-rsa AAAAB3Nza…省略…URmXD0= hadoop@slave2

$ chmod 644 authorized_keys

$ ssh -p2221 hadoop@10.0.24.16   $ ssh -p2221 slave1  # 分别测试ssh连通性

3、Java环境安装

### Hadoop集群均需安装Java环境

# mkdir /data  && cd /data

# tar zxf jdk-8u121-linux-x64.tar.gz

# ln -sv jdk1.8.0_121 jdk

# chown -R root. jdk*

# cat >> /etc/profile.d/java.sh<<‘EOF‘

# Set jave environment

export JAVA_HOME=/data/jdk

export CLASSPATH=.:$JAVA_HOME/lib:$JAVA_HOME/jre/lib

export PATH=$PATH:$JAVA_HOME/bin:$JAVA_HOME/jre/bin

EOF 

# source /etc/profile     # 及时生效     # java -version或# javac-version  # 查看版本信息

4、Hadoop集群安装

4.1 master上安装Hadoop

# cd /data

# hadoop-3.0.0-alpha2.tar.gz

# ln -sv hadoop-3.0.0-alpha2 hadoop   # mkdir -p /data/hadoop/logs # chown -Rhadoop:hadoop /data/hadoop/logs

# mkdir -p /data/hadoop/tmp          # 配置文件core-site.xml中配置使用

# mkdir -p /data/{hdfsname1,hdfsname2}/hdfs/name

# mkdir -p /data/{hdfsdata1,hdfsdata2}/hdfs/data

# chown -R hadoop:hadoop /data/hdfs*

# 以上四个文件目录hadfs-site.xml中配置使用

# cat >> /etc/profile.d/hadoop.sh<<‘EOF‘

# Set hadoop environment

export HADOOP_HOME=/data/hadoop

export PATH=$PATH:$HADOOP_HOME/bin

EOF 

# source /etc/profile

# chown -R hadoop:hadoop hadoop*

4.2 master上配置Hadoop

# cd /data/hadoop/etc/hadoop

4.2.1 hadoop-env.sh

# vim hadoop-env.sh       # master和slave末行均添加

# Set jave environment

export JAVA_HOME=/data/jdk

export HADOOP_SSH_OPTS="-p 2221"

4.2.2 core-site.xml

# vim core-site.xml

<configuration>

   <property>

       <name>fs.defaultFS</name>

       <value>hdfs://master:9000</value>

   </property>

   <property>

       <name>hadoop.tmp.dir</name>

       <value>/data/hadoop/tmp</value>

    </property>

   <property>

       <name>io.compression.codecs</name>

       <value>org.apache.hadoop.io.compress.DefaultCodec,com.hadoop.compression.lzo.LzoCodec,com.hadoop.compression.lzo.LzopCodec,org.apache.hadoop.io.compress.GzipCodec,org.apache

.hadoop.io.compress.BZip2Codec</value>

   </property>

   <property>

       <name>io.compression.codec.lzo.class</name>

       <value>com.hadoop.compression.lzo.LzoCodec</value>

   </property>

</configuration>

### 说明：

第1个<property>：定义hdfsnamenode的主机名和端口，本机，主机名在/etc/hosts设置

第2个<property>：定义如没有配置hadoop.tmp.dir参数，此时系统默认的临时目录为：/tmp/hadoo-hadoop。而这个目录在每次重启后都会被删掉，必须重新执行format才行，否则会出错。默认是NameNode、DataNode、JournalNode等存放数据的公共目录。用户也可以自己单独指定这三类节点的目录。这里的/data/hadoop/tmp目录与文件都是自己创建的，配置后在格式化namenode的时候也会自动创建。 

第3个<property>：定义hdfs使用压缩（本次测试暂时关闭了本项目，可以注释掉）

第4个<property>：定义压缩格式和解码器类（本次测试暂时关闭了本项目，可以注释掉）

4.2.3 hdfs-site.xml

# vim hdfs-site.xml

<configuration>

   <property>

       <name>dfs.name.dir</name>

       <value>file:///data/hdfsname1/hdfs/name,file:// /data/hdfsname2/hdfs/name</value>

       <description> </description>

   </property>

   <property>

       <name>dfs.data.dir</name>

       <value>file:///data/hdfsdata1/hdfs/data,file:///data/hdfsdata2/hdfs/data</value>

       <description> </description>

   </property>

   <property>

       <name>dfs.replication</name>

       <value>2</value>

   </property>

   <property>

       <name>dfs.datanode.du.reserved</name>

       <value>1073741824</value>

   </property>

   <property>

       <name>dfs.block.size</name>

       <value>134217728</value>

   </property>

   <property>

       <name>dfs.permissions</name>

       <value>false</value>

   </property>

</configuration>

第1个<property>：定义hdfs Namenode持久存储名字空间、事务日志路径。多路径可以使用“,”分割，这里配置模拟了多磁盘挂载。

第2个<property>：定义本地文件系统上DFS数据节点应存储其块的位置。可以逗号分隔目录列表，则数据将存储在所有命名的目录中，通常在不同的设备上。

第3个<property>：定义DataNode存储block的副本数量。默认值是3个，我们现在有2个 DataNode，该值不大2即可，份数越多越安全，但速度越慢。

第4个<property>：定义du操作返回。

第5个<property>：定义hdfs的存储块大小，默认64M，我用的128M。

第6个<property>：权限设置，最好不要。

4.2.4 mapred-site.xml

# cp -a mapred-site.xml.templatemapred-site.xml  

# vim mapred-site.xml

<configuration>

   <property>

       <name>mapreduce.framework.name</name>

       <value>yarn</value>

   </property>

   <property>

       <name>mapreduce.application.classpath</name>

       <value>

           /data/hadoop/etc/hadoop,

           /data/hadoop/share/hadoop/common/*,

           /data/hadoop/share/hadoop/common/lib/*,

           /data/hadoop/share/hadoop/hdfs/*,

           /data/hadoop/share/hadoop/hdfs/lib/*,

           /data/hadoop/share/hadoop/mapreduce/*,

           /data/hadoop/share/hadoop/mapreduce/lib/*,

           /data/hadoop/share/hadoop/yarn/*,

           /data/hadoop/share/hadoop/yarn/lib/*

     </value>

    </property>

</configuration>

###说明：

上面的mapreduce.application.classpath一开始没有配置，导致使用mapreduce时报错

Error: Could not find or load main classorg.apache.hadoop.mapreduce.v2.app.MRAppMaster

4.2.5 yarn-site.xml

# vim yarn-site.xml

<configuration>

 <property>

    <name>yarn.resourcemanager.hostname</name>

   <value>master</value>

 </property>

 <property>

   <name>yarn.nodemanager.aux-services</name>

   <value>mapreduce_shuffle</value>

 </property>

</configuration>

第1个<property>：定义指的是运行ResourceManager机器所在的节点.

第2个<property>：定义在hadoop2.2.0版本中是mapreduce_shuffle,一定要看清楚。

### 注意：本次测试使用了默认文件，没有添加任何内容。

4.2.6 workers

# vim workers      # 配置slave的主机名，否则slave节点不启动

slave1

slave2

4.3 slava上安装Hadoop

复制主节点master上的hadoop安装配置环境到所有的slave上，切记：目标路径要与master保持一致。

$ scp -P2221 hadoop.tar.gzhadoop@slave1:/home/hadoop

$ scp -P2221 hadoop.tar.gzhadoop@slave2:/home/hadoop

4.4配置防火墙

### 实验时可以关闭防火墙，避免不必要的麻烦，等后续陆续调试

4.5 Hadoop启动及其验证

4.5.1 master上格式化HDFS文件系统

### 注意回到master服务器上执行如下操作：

# su - hadoop

$ /data/hadoop/bin/hdfsnamenode -format      # 显示如下内容：

2017-03-15 19:02:50,062 INFO namenode.NameNode:STARTUP_MSG:

/************************************************************

STARTUP_MSG: Starting NameNode

STARTUP_MSG:  user = hadoop

STARTUP_MSG:  host = master/10.0.24.15

STARTUP_MSG:  args = [-format]

STARTUP_MSG:  version = 3.0.0-alpha2

……此处省略好多……

Re-format filesystem in Storage Directory/data/hdfsname1/hdfs/name ? (Y or N) y

Re-format filesystem in Storage Directory/data/hdfsname2/hdfs/name ? (Y or N) y

……此处省略好多……

2017-03-15 19:03:48,703 INFO namenode.FSImage:Allocated new BlockPoolId: BP-1344030132-10.0.24.15-1489575828688

……此处省略好多……

2017-03-15 19:03:48,999 INFO util.ExitUtil: Exitingwith status 0

2017-03-15 19:03:49,002 INFO namenode.NameNode:SHUTDOWN_MSG:

/************************************************************

SHUTDOWN_MSG: Shutting down NameNode atmaster/10.0.24.15

************************************************************/

4.5.2 启动校验停止集群

$ cd /data/hadoop/sbin   # master服务器上操作

（1）$ ./start-all.sh  # 启动       # 显示内容：WARNING WARN暂时没有解决，详见5、FAQ内

WARNING: Attempting to start all Apache Hadoopdaemons as hadoop in 10 seconds.

WARNING: This is not a recommended productiondeployment configuration.

WARNING: Use CTRL-C to abort.

Starting namenodes on [master]

Starting datanodes

Starting secondary namenodes [master]

2017-03-21 18:51:03,092 WARN util.NativeCodeLoader:Unable to load native-hadoop library for your platform... using builtin-javaclasses where applicable

Starting resourcemanager

Starting nodemanagers

（2）$ /data/jdk1.8.0_121/bin/jps       # master上查看进程

9058 SecondaryNameNode

9272 ResourceManager

9577 RunJar

8842 NameNode

9773 Jps

（3）$ /data/jdk1.8.0_121/bin/jps       # slave1\slave2上查看进程

5088 DataNode

5340 Jps

5213 NodeManager

（4）$ ./stop-all.sh                    # master服务器上操作停止集群

WARNING: Stopping all Apache Hadoop daemons as hadoopin 10 seconds.

WARNING: Use CTRL-C to abort.

Stopping namenodes on [master]

Stopping datanodes

Stopping secondary namenodes [master]

2017-03-21 18:57:20,746 WARN util.NativeCodeLoader:Unable to load native-hadoop library for your platform... using builtin-javaclasses where applicable

Stopping nodemanagers

slave1: WARNING: nodemanager did not stop gracefullyafter 5 seconds: Trying to kill with kill -9

slave2: WARNING: nodemanager did not stop gracefullyafter 5 seconds: Trying to kill with kill -9

Stopping resourcemanager

（5）$ /data/jdk1.8.0_121/bin/jps     # 再次查看进程都已经正常关闭

11500 Jps

（6）Web页面

1）http://192.168.24.15:8088

2）http://192.168.24.15:9870

4.5.3 Mapreduce程序测试

$ cd /data/hadoop/bin

1、第一种测试方法：

$ hadoop jar../share/hadoop/mapreduce/hadoop-mapreduce-examples-3.0.0-alpha2.jar pi 1 1  #说明成功

Number of Maps = 1

Samples per Map = 1

Wrote input for Map #0

Starting Job

2017-04-01 05:34:34,150 INFO client.RMProxy:Connecting to ResourceManager at master/192.168.24.15:8032

2017-04-01 05:34:35,765 INFO input.FileInputFormat:Total input files to process : 1

2017-04-01 05:34:35,876 INFO mapreduce.JobSubmitter:number of splits:1

2017-04-01 05:34:35,926 INFOConfiguration.deprecation:yarn.resourcemanager.system-metrics-publisher.enabled is deprecated. Instead,use yarn.system-metrics-publisher.enabled

2017-04-01 05:34:36,402 INFO mapreduce.JobSubmitter:Submitting tokens for job: job_1490957345671_0007

2017-04-01 05:34:36,939 INFO impl.YarnClientImpl:Submitted application application_1490957345671_0007

2017-04-01 05:34:37,085 INFO mapreduce.Job: The urlto track the job: http://master:8088/proxy/application_1490957345671_0007/

2017-04-01 05:34:37,086 INFO mapreduce.Job: Runningjob: job_1490957345671_0007

2017-04-01 05:34:47,336 INFO mapreduce.Job: Jobjob_1490957345671_0007 running in uber mode : false

2017-04-01 05:34:47,340 INFO mapreduce.Job:  map 0% reduce 0%

2017-04-01 05:34:57,496 INFO mapreduce.Job:  map 100% reduce 0%

2017-04-01 05:35:05,574 INFO mapreduce.Job:  map 100% reduce 100%

2017-04-01 05:35:05,588 INFO mapreduce.Job: Jobjob_1490957345671_0007 completed successfully

2、第二种测试方式：

（1）生成HDFS请求目录执行MapReduce任务

$ hdfs dfs -mkdir /user  

$ hdfs dfs -mkdir /user/hduser

（2）将输入文件拷贝到分布式文件系统

$ hdfs dfs -mkdir /user/hduser/input

$ hdfs dfs -put ../etc/hadoop/yarn-site.xml/user/hduser/input

（2）运行提供的示例程序

$ hadoop jar../share/hadoop/mapreduce/hadoop-mapreduce-examples-3.0.0-alpha2.jar grep/user/hduser/input output ‘dfs[a-z.]+‘……省略……

2017-03-31 10:58:46,650 INFO mapreduce.Job:  map 100% reduce 100%

2017-03-31 10:58:46,664 INFO mapreduce.Job: Jobjob_1490957345671_0003 completed successfully

2017-03-31 10:58:46,860 INFO mapreduce.Job: Counters:49

……省略……

http://192.168.24.15:9870里可以看到：

### 由于博客文字限制，只能分开写了：
Hadoop 3.0.0-alpha2安装（二）链接：

http://laowafang.blog.51cto.com/251518/1912345

刘政委 2017-04-01

本文出自 “老瓦房” 博客，请务必保留此出处http://laowafang.blog.51cto.com/251518/1912342

Hadoop 3.0.0-alpha2安装(一)