Hadoop伪分布式部署之yarn和mapreduce

前言

mapreduce是hadoop的分布式计算框架，它依赖于hadoop的分布式文件系统hdfs，关于hdfs的部署大家可以参考Hadoop伪分布式部署之hdfs。
mapreduce作为计算引擎，需要依赖于hadoop的分布式资源管理框架yarn，今天我们就来介绍一下yarn和mapreduce的伪分布式部署，相关环境如下：
操作系统：CentOS6.4
Java版本：Oracle jdk1.7
Hadoop版本：Hadoop2.5.0
主机hostname：hadoop01.datacenter.com
hadoop目录：/opt/modules/hadoop-2.5.0

1、Java环境设置

配置yarn的Java环境

[hadoop@hadoop01 ~]$ cd /opt/modules/hadoop-2.5.0/
[hadoop@hadoop01 hadoop-2.5.0]$ vim etc/hadoop/yarn-env.sh 
...
# some Java parameters
export JAVA_HOME=/opt/modules/jdk1.7.0_67
...
[hadoop@hadoop01 hadoop-2.5.0]$ 

配置mapreduce的Java环境

[hadoop@hadoop01 hadoop-2.5.0]$ vim etc/hadoop/mapred-env.sh 
...
export JAVA_HOME=/opt/modules/jdk1.7.0_67
...
[hadoop@hadoop01 hadoop-2.5.0]$ 

2、yarn和mapreduce参数设置

配置resourcemanager的hostname和nodemanager的附属服务

[hadoop@hadoop01 hadoop-2.5.0]$ vim etc/hadoop/yarn-site.xml 
...
<configuration>
    <property>
        <name>yarn.resourcemanager.hostname</name>
        <value>hadoop01.datacenter.com</value>
    </property>

    <property>
        <name>yarn.nodemanager.aux-services</name>
        <value>mapreduce_shuffle</value>
    </property>
</configuration>
...
[hadoop@hadoop01 hadoop-2.5.0]$ 

配置mapreduce运行环境为yarn

[hadoop@hadoop01 hadoop-2.5.0]$ cp etc/hadoop/mapred-site.xml.template etc/hadoop/mapred-site.xml
[hadoop@hadoop01 hadoop-2.5.0]$ vim etc/hadoop/mapred-site.xml
...
<configuration>
    <property>
        <name>mapreduce.framework.name</name>
        <value>yarn</value>
    </property>
</configuration>
...

3、启动hdfs和yarn服务

启动hdfs和yarn的相关服务

[hadoop@hadoop01 hadoop-2.5.0]$ sbin/hadoop-daemon.sh start namenode
starting namenode, logging to /opt/modules/hadoop-2.5.0/logs/hadoop-hadoop-namenode-hadoop01.datacenter.com.out
[hadoop@hadoop01 hadoop-2.5.0]$ sbin/hadoop-daemon.sh start datanode
starting datanode, logging to /opt/modules/hadoop-2.5.0/logs/hadoop-hadoop-datanode-hadoop01.datacenter.com.out
[hadoop@hadoop01 hadoop-2.5.0]$ sbin/yarn-daemon.sh start resourcemanager
starting resourcemanager, logging to /opt/modules/hadoop-2.5.0/logs/yarn-hadoop-resourcemanager-hadoop01.datacenter.com.out
[hadoop@hadoop01 hadoop-2.5.0]$ sbin/yarn-daemon.sh start nodemanager    
starting nodemanager, logging to /opt/modules/hadoop-2.5.0/logs/yarn-hadoop-nodemanager-hadoop01.datacenter.com.out
[hadoop@hadoop01 hadoop-2.5.0]$ 

通过jps命令检查服务启动是否成功

[hadoop@hadoop01 hadoop-2.5.0]$ jps
3557 NodeManager
3133 NameNode
3192 DataNode
3313 ResourceManager
3670 Jps
[hadoop@hadoop01 hadoop-2.5.0]$ 

通过8088端口访问yarn的webapp服务,我机器上的地址为http://hadoop01.datacenter.com:8088

4、wordcount测试

接下来我们使用官方的wordcount程序测试一下我们的环境。
首先我们在本地新建一个文本文档wordcount.txt，并输入一些单词

[hadoop@hadoop01 hadoop-2.5.0]$ vim /opt/datas/wordcount.txt
hadoop hdfs
yarn mapreduce
hbase spark
spark hello
[hadoop@hadoop01 hadoop-2.5.0]$ 

将wordcount.txt文件上传到hdfs的相对路径mapreduce/wordcount/input下

[hadoop@hadoop01 hadoop-2.5.0]$ bin/hdfs dfs -mkdir -p mapreduce/wordcount/input
18/04/07 22:10:05 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
[hadoop@hadoop01 hadoop-2.5.0]$ bin/hdfs dfs -put /opt/datas/wordcount.txt mapreduce/wordcount/input/
18/04/07 22:10:35 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
[hadoop@hadoop01 hadoop-2.5.0]$ 

使用bin/yarn执行官方自带的wordcount程序，输入为hdfs上mapreduce/wordcount/input目录，输出为hdfs上mapreduce/wordcount/output目录

[hadoop@hadoop01 hadoop-2.5.0]$ bin/yarn jar share/hadoop/mapreduce/hadoop-mapreduce-examples-2.5.0.jar wordcount mapreduce/wordcount/input mapreduce/wordcount/output
...
[hadoop@hadoop01 hadoop-2.5.0]$ 

程序完成之后我们可以查看hdfs的输出目录的内容

[hadoop@hadoop01 hadoop-2.5.0]$ bin/hdfs dfs -ls mapreduce/wordcount/output
18/04/07 22:19:54 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
Found 2 items
-rw-r--r--   1 hadoop supergroup          0 2018-04-07 22:17 mapreduce/wordcount/output/_SUCCESS
-rw-r--r--   1 hadoop supergroup         59 2018-04-07 22:17 mapreduce/wordcount/output/part-r-00000
[hadoop@hadoop01 hadoop-2.5.0]$ bin/hdfs dfs -cat mapreduce/wordcount/output/part*
18/04/07 22:20:16 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
hadoop  1
hbase   1
hdfs    1
hello   1
mapreduce       1
spark   2
yarn    1
[hadoop@hadoop01 hadoop-2.5.0]$ 

从上面的结果可以看出，输出文件已经对输入文件中的单词做了统计，并且是按照单词的字母顺序做的排序。

5、关闭hdfs和yarn服务

最后我们关闭hdfs和yarn服务

[hadoop@hadoop01 hadoop-2.5.0]$ sbin/hadoop-daemon.sh stop namenode
stopping namenode
[hadoop@hadoop01 hadoop-2.5.0]$ sbin/hadoop-daemon.sh stop datanode
stopping datanode
[hadoop@hadoop01 hadoop-2.5.0]$ sbin/yarn-daemon.sh stop resourcemanager
stopping resourcemanager
[hadoop@hadoop01 hadoop-2.5.0]$ sbin/yarn-daemon.sh stop nodemanager    
stopping nodemanager
[hadoop@hadoop01 hadoop-2.5.0]$

使用jps命令查看服务关闭是否成功

[hadoop@hadoop01 hadoop-2.5.0]$ jps
4738 Jps
[hadoop@hadoop01 hadoop-2.5.0]$ 

总结

1、我们的mapreduce跑在yarn上面，所以和yarn一起部署
2、首先配置mapreduce和yarn的java环境
3、接下来我们配置yarn的其他参数，这里我们分别平配置了resourcemanager的hostname和nodemanager的附属服务
4、最后我们配置了mapreduce运行在yarn上

注意事项：mapreduce的输出目录在程序开始之前必须不存在！