141.4. 分布式安装(CentOS 6 + hadoop-1.1.2)

HDFS:
      NameNode  ：管理节点
      DataNode  ：数据节点
      SecondaryNamenode : 数据源信息备份整理节点

MapReduce
JobTracker ：任务管理节点
Tasktracker ：任务运行节点

141.4.1. 准备工作

准备4台服务器，操作系统为 Centos 6.4 最小化安装

NameNode   192.168.2.10 hostname namenode
DataNode    192.168.2.11 hostname:datanode1
DataNode    192.168.2.12 hostname:datanode2

JobTracker 192.168.2.10 (也可单独配置一台,也可以与NameNode公用，这里只用到了HDFS，这台可有可无，准备上面4台即可)
TaskTracker (与DataNode共用)

设置网络使其可以互访，然后关闭防火墙与selinux

# yum update -y
# lokkit --disabled --selinux=disabled

Hadoop 重要的端口

1.Job Tracker 管理界面： 50030
2.HDFS 管理界面： 50070
3.HDFS通信端口： 9000
4.MapReduce通信端口： 9001

过程 141.3. Hadoop - 准备工作

为所有服务器安装Java运行环境

以 CentOS 6.4 为例
```
# yum install java-1.7.0-openjdk
```

在所有服务器上安装 Hadoop

安装方案有下面两种 RPM与YUM，选择其中一种

# rpm -ivh http://ftp.cuhk.edu.hk/pub/packages/apache.org/hadoop/common/hadoop-1.1.2/hadoop-1.1.2-1.x86_64.rpm
Retrieving http://ftp.cuhk.edu.hk/pub/packages/apache.org/hadoop/common/hadoop-1.1.2/hadoop-1.1.2-1.x86_64.rpm
Preparing...                ########################################### [100%]1:hadoop                 ########################################### [100%]

yum localinstall http://ftp.cuhk.edu.hk/pub/packages/apache.org/hadoop/common/hadoop-1.1.2/hadoop-1.1.2-1.x86_64.rpm

如果网络比较慢，可以使用Wget或axel下载后安装

wget http://ftp.cuhk.edu.hk/pub/packages/apache.org/hadoop/common/hadoop-1.1.2/hadoop-1.1.2-1.x86_64.rpm
yum localinstall hadoop-1.1.2-1.x86_64.rpm

Hadoop 用户

# cat /etc/passwd | grep Hadoop
mapred:x:202:123:Hadoop MapReduce:/tmp:/bin/bash
hdfs:x:201:123:Hadoop HDFS:/tmp:/bin/bash

配置/etc/hosts文件

cat >> /etc/hosts <<EOD###############################
# Hadoop Host
###############################
#NameNode
192.168.2.10    namenode.example.com#DataNode
192.168.2.11    datanode1.example.com
192.168.2.12    datanode2.example.comEOD

生成其密钥

# ssh-keygen
Generating public/private rsa key pair.
Enter file in which to save the key (/root/.ssh/id_rsa):
Enter passphrase (empty for no passphrase):
Enter same passphrase again:
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:
cc:6f:30:76:82:28:96:13:c8:e6:bc:d7:5b:2d:11:d7 root@images-upload
The key's randomart image is:
+--[ RSA 2048]----+
|                 |
|..        .      |
|.o.    . . E     |
|+  o . +o        |
| o= . ..S .      |
| ..o.  .o*       |
| . . . o .o      |
|  .   o ..       |
|     .           |
+-----------------+

植入公钥证书

向DataNode节点所有的服务器植入公钥证书

ssh-copy-id root@datanode1.example.com
ssh-copy-id root@datanode2.example.com

只需要输入yes后，再输入密码即可完成公钥证书的植入。过程类似下面：

# ssh-copy-id root@datanode1.example.com
The authenticity of host 'datanode1.example.com (192.168.2.11)' can't be established.
RSA key fingerprint is f1:0b:b1:63:1a:f6:ac:a3:da:4f:14:b5:f0:cc:df:67.
Are you sure you want to continue connecting (yes/no)? yes 输入yes
Warning: Permanently added 'datanode1.example.com' (RSA) to the list of known hosts.
root@datanode1.example.com's password: 输入密码
Now try logging into the machine, with "ssh 'root@datanode1.example.com'", and check in:.ssh/authorized_keysto make sure we haven't added extra keys that you weren't expecting.# ssh-copy-id root@datanode2.example.com
The authenticity of host 'datanode2.example.com (192.168.2.12)' can't be established.
RSA key fingerprint is f1:0b:b1:63:1a:f6:ac:a3:da:4f:14:b5:f0:cc:df:67.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added 'datanode2.example.com,192.168.2.12' (RSA) to the list of known hosts.
root@datanode2.example.com's password:
Now try logging into the machine, with "ssh 'root@datanode2.example.com'", and check in:.ssh/authorized_keysto make sure we haven't added extra keys that you weren't expecting.

完成后测试登陆，如果没有提示密码直接进入表示正确

# ssh root@datanode1.example.com
# exit

141.4.2. NameNode 配置名称节点

配置文件

core-site.xml   common属性配置
hdfs-site.xml    HDFS属性配置
mapred-site.xml  MapReduce属性配置
hadoop-env.sh    hadooop 环境变量配置

过程 141.4. Hadoop - NameNode

配置文件 hadoop-env.sh

将 /usr/java/default 改为 /usr

# cp hadoop-env.sh hadoop-env.sh.original
# sed -i "s:/usr/java/default:/usr:" hadoop-env.sh

配置文件 core-site.xml

# cp core-site.xml core-site.xml.original<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?><!-- Put site-specific property overrides in this file. --><configuration><property><name>fs.default.name</name><value>hdfs://namenode.example.com:9000</value></property><property><name>hadoop.tmp.dir</name><value>/var/tmp/hadoop</value></property>
</configuration>

fs.default.name: NameNode的URI。hdfs://主机名:端口/

hadoop.tmp.dir: Hadoop的默认临时路径，

配置文件 mapred-site.xml

# cp mapred-site.xml mapred-site.xml.original
<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?><!-- Put site-specific property overrides in this file. --><configuration><property><name>mapred.job.tracker</name><value>namenode.example.com:9001</value></property><property><name>mapred.local.dir</name><value>/var/tmp/hadoop</value></property>
</configuration>

mapred.job.tracker: JobTracker的主机和端口。

配置文件 hdfs-site.xml

# cp hdfs-site.xml hdfs-site.xml.original<?xml version="1.0"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?><!-- Put site-specific property overrides in this file. --><configuration><property><name>dfs.name.dir</name><value>/var/hadoop/name1</value><description>  </description></property><property><name>dfs.data.dir</name><value>/var/hadoop/hdfs/data1</value><description> </description></property><property><name>dfs.replication</name><value>2</value></property>
</configuration>

dfs.name.dir: NameNode持久存储名字空间及事务日志的本地文件系统路径。 当这个值是一个逗号分割的目录列表时，nametable数据将会被复制到所有目录中做冗余备份。
2）   dfs.data.dir是DataNode存放块数据的本地文件系统路径，逗号分割的列表。 当这个值是逗号分割的目录列表时，数据将被存储在所有目录下，通常分布在不同设备上。
3）dfs.replication是数据需要备份的数量，默认是3，如果此数大于集群的机器数会出错。

配置masters和slaves主从结点

备份masters与slaves配置文件

 cp masters masters.originalcp slaves slaves.original

cat > /etc/hadoop/masters <<EOD
namenode.example.com
EOD

cat > /etc/hadoop/slaves <<EOD
datanode1.example.com
datanode2.example.com
EOD

复制配置文件

# cd /etc/hadoop/
# scp hadoop-env.sh core-site.xml mapred-site.xml hdfs-site.xml masters slaves root@datanode1.example.com:/etc/hadoop/
# scp hadoop-env.sh core-site.xml mapred-site.xml hdfs-site.xml masters slaves root@datanode2.example.com:/etc/hadoop/

控制台输出类似下面表示复制成功。

# scp hadoop-env.sh core-site.xml mapred-site.xml hdfs-site.xml masters slaves root@datanode1.example.com:/etc/hadoop/
hadoop-env.sh                                                                          100% 2116     2.1KB/s   00:00
core-site.xml                                                                          100%  412     0.4KB/s   00:00
mapred-site.xml                                                                        100%  406     0.4KB/s   00:00
hdfs-site.xml                                                                          100%  595     0.6KB/s   00:00
masters                                                                                100%   21     0.0KB/s   00:00
slaves

将 NameNode 上的配置文件复制给 DataNode

启动 Hadoop

创建工作目录

# mkdir /var/hadoop/
# mkdir /var/hadoop/name{1,2}
# su - hdfs -c  "mkdir -p  /var/hadoop/hdfs/data{1,2}"

# hadoop namenode -format
13/04/23 14:35:33 INFO namenode.NameNode: STARTUP_MSG:
/************************************************************
STARTUP_MSG: Starting NameNode
STARTUP_MSG:   host = namenode.example.com/192.168.2.10
STARTUP_MSG:   args = [-format]
STARTUP_MSG:   version = 1.1.2
STARTUP_MSG:   build = https://svn.apache.org/repos/asf/hadoop/common/branches/branch-1.1 -r 1440782; compiled by 'hortonfo' on Thu Jan 31 02:06:43 UTC 2013
************************************************************/
Re-format filesystem in /var/hadoop/name1 ? (Y or N) Y
13/04/23 14:35:37 INFO util.GSet: VM type       = 64-bit
13/04/23 14:35:37 INFO util.GSet: 2% max memory = 2.475 MB
13/04/23 14:35:37 INFO util.GSet: capacity      = 2^18 = 262144 entries
13/04/23 14:35:37 INFO util.GSet: recommended=262144, actual=262144
13/04/23 14:35:37 INFO namenode.FSNamesystem: fsOwner=root
13/04/23 14:35:37 INFO namenode.FSNamesystem: supergroup=supergroup
13/04/23 14:35:37 INFO namenode.FSNamesystem: isPermissionEnabled=true
13/04/23 14:35:37 INFO namenode.FSNamesystem: dfs.block.invalidate.limit=100
13/04/23 14:35:37 INFO namenode.FSNamesystem: isAccessTokenEnabled=false accessKeyUpdateInterval=0 min(s), accessTokenLifetime=0 min(s)
13/04/23 14:35:38 INFO namenode.NameNode: Caching file names occuring more than 10 times
13/04/23 14:35:38 INFO common.Storage: Image file of size 110 saved in 0 seconds.
13/04/23 14:35:38 INFO namenode.FSEditLog: closing edit log: position=4, editlog=/var/hadoop/name1/current/edits
13/04/23 14:35:38 INFO namenode.FSEditLog: close success: truncate to 4, editlog=/var/hadoop/name1/current/edits
13/04/23 14:35:38 INFO common.Storage: Storage directory /var/hadoop/name1 has been successfully formatted.
13/04/23 14:35:38 INFO common.Storage: Image file of size 110 saved in 0 seconds.
13/04/23 14:35:38 INFO namenode.FSEditLog: closing edit log: position=4, editlog= /var/hadoop/name2/current/edits
13/04/23 14:35:38 INFO namenode.FSEditLog: close success: truncate to 4, editlog= /var/hadoop/name2/current/edits
13/04/23 14:35:38 INFO common.Storage: Storage directory  /var/hadoop/name2 has been successfully formatted.
13/04/23 14:35:38 INFO namenode.NameNode: SHUTDOWN_MSG:
/************************************************************
SHUTDOWN_MSG: Shutting down NameNode at namenode.example.com/192.168.2.10
************************************************************/

# chown hdfs:hadoop -R /var/hadoop

# /etc/init.d/hadoop-namenode start
# /etc/init.d/hadoop-datanode start

http://192.168.2.10:50070/

141.4.3. DataNode 配置数据节点

过程 141.5. Hadoop - DataNode

创建hadoop数据存储目录

mkdir /var/hadoop/
chown hdfs:hadoop -R /var/hadoop
su - hdfs -c  "mkdir -p  /var/hadoop/hdfs/data1"

启动 Hadoop
```
# /etc/init.d/hadoop-datanode start
```

141.4.4. Hadoop UI (WEB界面)

常用访问页面

1. HDFS 界面
http://hostname:50070
2. MapReduce 管理界面
http://hostname:50030

141.4.5. 测试Hadoop

将install.log文件拷贝到分布式文件系统

hadoop fs -mkdir test
hadoop fs -put install.log test

显示文件内容

# hadoop dfs -cat test/install.log

查看目录结构

# hadoop dfs -ls
Found 1 items
drwxr-xr-x   - root supergroup          0 2013-04-23 15:20 /user/root/test
[root@namenode ~]# hadoop dfs -ls test
Found 1 items
-rw-r--r--   2 root supergroup      10278 2013-04-23 15:20 /user/root/test/install.log

原文出处：Netkiller 系列手札
本文作者：陈景峯
转载请与作者联系，同时请务必标明文章原始出处和作者信息及本声明。