OpenStack环境下Hadoop2.2.0环境搭建

OpenStack目前已经成为众多云计算厂商搭建私有云的首选，众多学术机构也使用OpenStack搭建小规模测试环境供学生实验，在此分享使用OpenStack虚拟机搭建Hadoop2.2.0环境的过程。

1.VM环境准备

OpenStack版本：Folsom

a. 发起三台测试虚拟机，操作系统为 Ubuntu-12.04.2-x86_64

b. 配置IP地址，因为在F版本的OpenStack中，网络采用FlatDHCP模式使得虚拟机获得10.0.x.x段的Fixed IP地址，因此需要在虚拟机中配置/etc/hosts文件。

# vim /etc/hosts

127.0.0.1 localhost localhost.localdomain
10.0.0.225 hdp-server-01
10.0.1.19 hdp-server-02
10.0.1.17 hdp-server-03

c. 用root在每台机器上新建用户 yarn，使用同样的密码

# useradd -m -s /bin/bash yarn
# passwd yarn
Enter new UNIX password: 
Retype new UNIX password: 
passwd: password updated successfully

d. 设置ssh无密码互访

#每台机器
$ su yarn
$ cd ~
$ ssh-keygen -t rsa
$ cat .ssh/id_rsa.pub >> .ssh/authorized_keys
#可以使用 ssh localhost 测试是否可以无密码访问
#相互之间可以将.ssh/authorized_keys的内容互拷到对方的.ssh/authorized_keys文件中。

e. 使用yarn账户，通过/etc/hosts文件中填写的主机名进行互访，并验证是否无密码登录。

因为采用64位的操作系统，因此不能够直接使用从官网下载的文件进行安装，必须手动编译。以下为编译过程：

2.编译Hadoop2.2.0

a. 配置JDK环境变量，假设jdk文件夹为/usr/java/jdk1.7.0_45

su yarn
# vim ~/.bashrc
# 追加写入

export JAVA_HOME=/usr/local/java/jdk1.7.0_45
export JRE_HOME=${JAVA_HOME}/jre
export CLASSPATH=.:${JAVA_HOME}/lib:${JRE_HOME}/lib
export PATH=${JAVA_HOME}/bin:$PATH

$source ~/.bashrc #使设置生效

b. 安装编译所需依赖

$sudo apt-get install g++ autoconf automake libtool cmake zlib1g-dev pkg-config libssl-dev

c. 安装protobuf 2.5.0

$cd $HOME/protobuf2.5.0

$./configure --prefix=/usr
$sudo make
$sudo make check
$sudo make install

$ protoc --version
libprotoc 2.5.0

d. 安装maven

$ sudo apt-get install maven

e. 开始编译

$ cd ~
$ tar -xzvf hadoop-2.2.0-src.tar.gz
$ cd hadoop-2.2.0-src/
$ mvn package -Pdist,native -DskipTests -Dtar

编译大概耗时约30分钟， 编译完后的文件在  hadoop-2.2.0-src/hadoop-dist/target 路径中。如果发起的虚拟机都是相同操作系统，编译只需要在一台机器上执行。

验证编译结果：

yarn@hdp-server-01:~$ $HOME/hadoop-2.2.0/bin/hadoop version
Hadoop 2.2.0
Subversion Unknown -r Unknown
Compiled by yarn on 2013-11-05T06:41Z
Compiled with protoc 2.5.0
From source with checksum 79e53ce7994d1628b240f09af91e1af4
This command was run using /home/yarn/hadoop-2.2.0/share/hadoop/common/hadoop-common-2.2.0.jar

yarn@hdp-server-01:~$ file $HOME/hadoop-2.2.0/lib/native/*
/home/yarn/hadoop-2.2.0/lib/native/libhadoop.a:        current ar archive
/home/yarn/hadoop-2.2.0/lib/native/libhadooppipes.a:   current ar archive
/home/yarn/hadoop-2.2.0/lib/native/libhadoop.so:       ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, BuildID[sha1]=0xaa74c9d23bfe750f160412e4465b14c88cf1c650, not stripped
/home/yarn/hadoop-2.2.0/lib/native/libhadoop.so.1.0.0: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, BuildID[sha1]=0xaa74c9d23bfe750f160412e4465b14c88cf1c650, not stripped
/home/yarn/hadoop-2.2.0/lib/native/libhadooputils.a:   current ar archive
/home/yarn/hadoop-2.2.0/lib/native/libhdfs.a:          current ar archive
/home/yarn/hadoop-2.2.0/lib/native/libhdfs.so:         ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, BuildID[sha1]=0x89671252f3c5fb7034425e80c9d31ea67da75c4d, not stripped
/home/yarn/hadoop-2.2.0/lib/native/libhdfs.so.0.0.0:   ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, BuildID[sha1]=0x89671252f3c5fb7034425e80c9d31ea67da75c4d, not stripped

3. 安装配置Hadoop2.2.0

假定各node角色划分如下：

hdp-server-01     resourcemanager, nodemanager, proxyserver,historyserver, datanode, namenode
hdp-server-02     datanode, nodemanager
hdp-server-03     datanode, nodemanager

a. 目录准备

mkdir -p ~/yarn_data/tmp
mkdir -p ~/yarn_data/mapred

并将编译好的hadoop-2.2.0文件夹拷贝至用户$HOME

b. 配置环境变量（追加至~/.bashrc）

#hadoop env
export HADOOP_HOME="$HOME/hadoop-2.2.0"
export HADOOP_PREFIX="$HADOOP_HOME/"
export YARN_HOME=$HADOOP_HOME
export HADOOP_MAPRED_HOME="$HADOOP_HOME"
export HADOOP_COMMON_HOME="$HADOOP_HOME"
export HADOOP_HDFS_HOME="$HADOOP_HOME"
export HADOOP_CONF_DIR="$HADOOP_HOME/etc/hadoop/"
export YARN_CONF_DIR=$HADOOP_CONF_DIR
export PATH="$HADOOP_HOME/bin:$HADOOP_HOME/sbin:$PATH"

c. 修改官方启动脚本 

这个步骤是参考http://www.cnblogs.com/lucius/p/3435296.html，作者指出这是一个bug，本人修改后测试可以运行。

$ cd $YARN_HOME/libexec/
$ vim hadoop-config.sh
#修改第96行代码为：
export HADOOP_SLAVES="${HADOOP_CONF_DIR}/$1"
#保存退出vim

d. 设置配置文件

  
    fs.defaultFS
    hdfs://hdp-server-01:9000
  
  
    hadoop.tmp.dir
    /home/yarn/yarn_data/tmp/hadoop-grid
  

  
    dfs.replication
    3
  

   
     yarn.nodemanager.aux-services
     mapreduce_shuffle
  
  
     yarn.resourcemanager.address
     hdp-server-01:8032
  
  
      yarn.resourcemanager.resource-tracker.address
      hdp-server-01:8031
  
  
      yarn.resourcemanager.admin.address
      hdp-server-01:8033
  
  
      yarn.resourcemanager.scheduler.address
      hdp-server-01:8030
  
  
      yarn.nodemanager.loacl-dirs
      /home/yarn/yarn_data/mapred/nodemanager
      true
  
  
      yarn.web-proxy.address
      hdp-server-01:8888
  
  
     yarn.nodemanager.aux-services.mapreduce.shuffle.class
     org.apache.hadoop.mapred.ShuffleHandler
  

  
    mapreduce.framework.name
    yarn
  

至此，配置文件修改完毕，修改完后将 $HOME/hadoop-2.2.0 及 $HOME/yarn_data 两个文件分别拷贝至其他机器的同样位置，注意确保文件所有者为yarn。

e. HDFS格式化

$ hdfs namenode -format

f. 启动hadoop

@hdp-server-01 #在不同vm上启动的服务不同，根据划分的角色

$cd $YARN_HOME
$sbin/hadoop-daemon.sh  --script hdfs start namenode　　# 启动namenode
$sbin/hadoop-daemon.sh --script hdfs start datanode　　# 启动datanode
$sbin/yarndaemon.shstart nodemanager　　#启动nodemanager
$sbin/yarn-daemon.sh   start resourcemanager　# 启动resourcemanager
$sbin/yarn-daemon.shstart proxyserver　　#启动web App proxy
$sbin/mr-jobhistory-daemon.sh   start historyserver

jps查看
$ jps
8770 ResourceManager
11609 Jps
8644 NodeManager
9071 JobHistoryServer
8479 NameNode
9000 WebAppProxyServer
8552 DataNode

@hdp-server-02
@hdp-server-03

$cd $YARN_HOME
$sbin/yarndaemon.shstart nodemanager　　# 启动nodemanager
$sbin/hadoop-daemon.sh  --script hdfs start datanode　　# 启动datanode

jps查看
$ jps
6691 NodeManager
9089 Jps
6787 DataNode

至此，集群搭建完毕，跑一个测试用例试试：

cd $YARN_HOME
$ bin/yarn jar share/hadoop/mapreduce/hadoop-mapreduce-examples-2.2.0.jar pi 10 1000

这是mongodb蒙特卡洛算法计算圆周率的测试用例，pi后跟的两个数字分别表示使用多少个map以及计算的精度。结果如下：

Number of Maps  = 10
Samples per Map = 1000
Wrote input for Map #0
Wrote input for Map #1
Wrote input for Map #2
Wrote input for Map #3
Wrote input for Map #4
Wrote input for Map #5
Wrote input for Map #6
Wrote input for Map #7
Wrote input for Map #8
Wrote input for Map #9
Starting Job
13/12/22 17:50:42 INFO client.RMProxy: Connecting to ResourceManager at hdp-server-01/10.0.0.225:8032
13/12/22 17:50:43 INFO input.FileInputFormat: Total input paths to process : 10
13/12/22 17:50:43 INFO mapreduce.JobSubmitter: number of splits:10
13/12/22 17:50:43 INFO Configuration.deprecation: user.name is deprecated. Instead, use mapreduce.job.user.name
13/12/22 17:50:43 INFO Configuration.deprecation: mapred.jar is deprecated. Instead, use mapreduce.job.jar
13/12/22 17:50:43 INFO Configuration.deprecation: mapred.map.tasks.speculative.execution is deprecated. Instead, use mapreduce.map.speculative
13/12/22 17:50:43 INFO Configuration.deprecation: mapred.reduce.tasks is deprecated. Instead, use mapreduce.job.reduces
13/12/22 17:50:43 INFO Configuration.deprecation: mapred.output.value.class is deprecated. Instead, use mapreduce.job.output.value.class
13/12/22 17:50:43 INFO Configuration.deprecation: mapred.reduce.tasks.speculative.execution is deprecated. Instead, use mapreduce.reduce.speculative
13/12/22 17:50:43 INFO Configuration.deprecation: mapreduce.map.class is deprecated. Instead, use mapreduce.job.map.class
13/12/22 17:50:43 INFO Configuration.deprecation: mapred.job.name is deprecated. Instead, use mapreduce.job.name
13/12/22 17:50:43 INFO Configuration.deprecation: mapreduce.reduce.class is deprecated. Instead, use mapreduce.job.reduce.class
13/12/22 17:50:43 INFO Configuration.deprecation: mapreduce.inputformat.class is deprecated. Instead, use mapreduce.job.inputformat.class
13/12/22 17:50:43 INFO Configuration.deprecation: mapred.input.dir is deprecated. Instead, use mapreduce.input.fileinputformat.inputdir
13/12/22 17:50:43 INFO Configuration.deprecation: mapred.output.dir is deprecated. Instead, use mapreduce.output.fileoutputformat.outputdir
13/12/22 17:50:43 INFO Configuration.deprecation: mapreduce.outputformat.class is deprecated. Instead, use mapreduce.job.outputformat.class
13/12/22 17:50:43 INFO Configuration.deprecation: mapred.map.tasks is deprecated. Instead, use mapreduce.job.maps
13/12/22 17:50:43 INFO Configuration.deprecation: mapred.output.key.class is deprecated. Instead, use mapreduce.job.output.key.class
13/12/22 17:50:43 INFO Configuration.deprecation: mapred.working.dir is deprecated. Instead, use mapreduce.job.working.dir
13/12/22 17:50:44 INFO mapreduce.JobSubmitter: Submitting tokens for job: job_1387700249346_0004
13/12/22 17:50:44 INFO impl.YarnClientImpl: Submitted application application_1387700249346_0004 to ResourceManager at hdp-server-01/10.0.0.225:8032
13/12/22 17:50:44 INFO mapreduce.Job: The url to track the job: http://hdp-server-01:8888/proxy/application_1387700249346_0004/
13/12/22 17:50:44 INFO mapreduce.Job: Running job: job_1387700249346_0004
13/12/22 17:50:53 INFO mapreduce.Job: Job job_1387700249346_0004 running in uber mode : false
13/12/22 17:50:53 INFO mapreduce.Job:  map 0% reduce 0%
13/12/22 17:51:03 INFO mapreduce.Job:  map 40% reduce 0%
13/12/22 17:51:13 INFO mapreduce.Job:  map 90% reduce 0%
13/12/22 17:51:14 INFO mapreduce.Job:  map 100% reduce 0%
13/12/22 17:51:15 INFO mapreduce.Job:  map 100% reduce 100%
13/12/22 17:51:16 INFO mapreduce.Job: Job job_1387700249346_0004 completed successfully
13/12/22 17:51:16 INFO mapreduce.Job: Counters: 43
	File System Counters
		FILE: Number of bytes read=226
		FILE: Number of bytes written=878638
		FILE: Number of read operations=0
		FILE: Number of large read operations=0
		FILE: Number of write operations=0
		HDFS: Number of bytes read=2680
		HDFS: Number of bytes written=215
		HDFS: Number of read operations=43
		HDFS: Number of large read operations=0
		HDFS: Number of write operations=3
	Job Counters 
		Launched map tasks=10
		Launched reduce tasks=1
		Data-local map tasks=10
		Total time spent by all maps in occupied slots (ms)=142127
		Total time spent by all reduces in occupied slots (ms)=8333
	Map-Reduce Framework
		Map input records=10
		Map output records=20
		Map output bytes=180
		Map output materialized bytes=280
		Input split bytes=1500
		Combine input records=0
		Combine output records=0
		Reduce input groups=2
		Reduce shuffle bytes=280
		Reduce input records=20
		Reduce output records=0
		Spilled Records=40
		Shuffled Maps =10
		Failed Shuffles=0
		Merged Map outputs=10
		GC time elapsed (ms)=2606
		CPU time spent (ms)=11090
		Physical memory (bytes) snapshot=2605563904
		Virtual memory (bytes) snapshot=11336945664
		Total committed heap usage (bytes)=2184183808
	Shuffle Errors
		BAD_ID=0
		CONNECTION=0
		IO_ERROR=0
		WRONG_LENGTH=0
		WRONG_MAP=0
		WRONG_REDUCE=0
	File Input Format Counters 
		Bytes Read=1180
	File Output Format Counters 
		Bytes Written=97
Job Finished in 34.098 seconds
Estimated value of Pi is 3.14080000000000000000

总结：

1. 在OpenStack环境启动的虚拟机中搭建Hadoop与物理机搭建并无太大不同，但需要注意虚拟机获取到的IP地址，用openstack分配的浮动ip（Floating ip）往往不能使用。因为浮动ip是由nova-network设置，用于nat转发的，虚拟机自身并不知道这个地址。

2. 集群中使用的虚拟机最好是同样的操作系统，这样可以使用编译好的文件，因为在Hadoop2.2.0框架中 hdfs不存在Master节点，因此每个节点的配置文件都是相同的，故可以先发起一台虚拟机，安装配置完之后将其做成镜像，后续可以起多个节点，区别在于启动的服务不同。

参考：http://www.cnblogs.com/lucius/p/3435296.html