安装hadoop集群(MultiCluster)-创新互联

配置环境

本文档安装hadoop集群环境，一个master作为namenode节点，一个slave作为datanode节点：

(1) master：

os: CentOS release 6.5 (Final)

ip: 172.16.101.58

user：root

hadoop-2.9.0.tar.gz

(2) slave：

os: CentOS release 6.5 (Final)

ip: 172.16.101.59

user：root

hadoop-2.9.0.tar.gz

前提条件

(1) master和slave都安装好java环境，并配置好环境变量；

(2)master节点解压好hadoop-2.9.0.tar.gz，并配置好环境变量；

(3)本篇文档使用的是root用户安装，所以需要master上的root用户可以ssh无密码使用root用户登录slave节点；

配置集群文件

在 master节点上执行（本文档先在master节点上配置文件，然后通过scp拷贝到其他slave节点）

（1）slaves文件：将作为 DataNode 的主机名或者ip写入该文件，每行一个，默认为 localhost，所以在伪分布式配置时，节点既作为 NameNode 也作为 DataNode。

[root@sht-sgmhadoopdn-01 hadoop]# cat slaves

172.16.101.59

（2）文件core-site.xml

[root@sht-sgmhadoopdn-01 hadoop]#cat /usr/local/hadoop-2.9.0/etc/hadoop/core-site.xml

        fs.defaultFS

        hdfs://172.16.101.58:9000

        hadoop.tmp.dir

        /usr/local/hadoop-2.9.0/tmp

        Abase for other temporary directories.

（3）文件hdfs-site.xml

[root@sht-sgmhadoopdn-01 hadoop]# cat /usr/local/hadoop-2.9.0/etc/hadoop/hdfs-site.xml

                dfs.namenode.secondary.http-address

                172.16.101.58:50090

                dfs.replication

                1

                dfs.namenode.name.dir

                file:/usr/local/hadoop-2.9.0/tmp/dfs/name

                dfs.datanode.data.dir

                file:/usr/local/hadoop-2.9.0/tmp/dfs/data

（4）文件mapred-site.xml

[root@sht-sgmhadoopdn-01 hadoop]# cat /usr/local/hadoop-2.9.0/etc/hadoop/mapred-site.xml

    mapreduce.framework.name

    yarn

        mapreduce.jobhistory.address

        172.16.101.58:10020

        mapreduce.jobhistory.webapp.address

        172.16.101.58:19888

（5）文件yarn-site.xml

[root@sht-sgmhadoopdn-01 hadoop]# cat /usr/local/hadoop-2.9.0/etc/yarn-site.xml

        yarn.resourcemanager.hostname

        172.16.101.58

        yarn.nodemanager.aux-services

        mapreduce_shuffle

配置好后，将 Master上的 /usr/local/hadoop-2.9.0文件复制到各个节点上。因为之前有跑过伪分布式模式，建议在切换到集群模式前先删除之前的临时文件。

[root@sht-sgmhadoopdn-01 local]# rm -rf ./hadoop-2.9.0/tmp

[root@sht-sgmhadoopdn-01 local]# rm -rf ./hadoop-2.9.0/logs

[root@sht-sgmhadoopdn-01 local]# tar -zcf  hadoop-2.9.0.master.tar.gz  /usr/local/hadoop-2.9.0

[root@sht-sgmhadoopdn-01 local]# scp hadoop-2.9.0.master.tar.gz sht-sgmhadoopdn-02:/usr/local/

在 Slave节点上执行

[root@sht-sgmhadoopdn-02 local]# tar -zxf hadoop-2.9.0.master.tar.gz

启动hadoop集群

在 master节点上执行：

#第一次启动需要格式化HDFS,以后再启动不需要

[root@sht-sgmhadoopdn-01 hadoop-2.9.0]# hdfs namenode -format

[root@sht-sgmhadoopdn-01 hadoop-2.9.0]# start-dfs.sh

[root@sht-sgmhadoopdn-01 hadoop-2.9.0]# start-yarn.sh

[root@sht-sgmhadoopdn-01 hadoop-2.9.0]# mr-jobhistory-daemon.sh start historyserver

[root@sht-sgmhadoopdn-01 hadoop-2.9.0]# jps

20289 JobHistoryServer

19730 ResourceManager

18934 NameNode

19163 SecondaryNameNode

20366 Jps

在 Slave节点上执行：

[root@sht-sgmhadoopdn-02 hadoop]# jps

32147 DataNode

535 Jps

32559 NodeManager

在 master节点上执行：

[root@sht-sgmhadoopdn-01 hadoop]# hdfs dfsadmin -report

Configured Capacity: 75831140352 (70.62 GB)

Present Capacity: 21246287872 (19.79 GB)

DFS Remaining: 21246263296 (19.79 GB)

DFS Used: 24576 (24 KB)

DFS Used%: 0.00%

Under replicated blocks: 0

Blocks with corrupt replicas: 0

Missing blocks: 0

Missing blocks (with replication factor 1): 0

Pending deletion blocks: 0

-------------------------------------------------

Live datanodes (1):                               #存活的slave数量

Name: 172.16.101.59:50010 (sht-sgmhadoopdn-02)

Hostname: sht-sgmhadoopdn-02

Decommission Status : Normal

Configured Capacity: 75831140352 (70.62 GB)

DFS Used: 24576 (24 KB)

Non DFS Used: 50732867584 (47.25 GB)

DFS Remaining: 21246263296 (19.79 GB)

DFS Used%: 0.00%

DFS Remaining%: 28.02%

Configured Cache Capacity: 0 (0 B)

Cache Used: 0 (0 B)

Cache Remaining: 0 (0 B)

Cache Used%: 100.00%

Cache Remaining%: 0.00%

Xceivers: 1

Last contact: Wed Dec 27 11:08:46 CST 2017

Last Block Report: Wed Dec 27 11:02:01 CST 2017

Console管理平台

NameNodehttp://172.16.101.58:50070

执行分布式实例MapReduce Job

[root@sht-sgmhadoopdn-01 hadoop]# hdfs dfs -mkdir -p /user/root/input

[root@sht-sgmhadoopdn-01 hadoop]# hdfs dfs -put /usr/local/hadoop-2.9.0/etc/hadoop/*.xml  input

[root@sht-sgmhadoopdn-01 hadoop]# hadoop jar /usr/local/hadoop-2.9.0/share/hadoop/mapreduce/hadoop-mapreduce-examples-2.9.0.jar grep input output 'dfs[a-z.]+'

17/12/27 11:25:33 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable

17/12/27 11:25:34 INFO client.RMProxy: Connecting to ResourceManager at /172.16.101.58:8032

17/12/27 11:25:36 INFO input.FileInputFormat: Total input files to process : 9

17/12/27 11:25:36 INFO mapreduce.JobSubmitter: number of splits:9

17/12/27 11:25:37 INFO Configuration.deprecation: yarn.resourcemanager.system-metrics-publisher.enabled is deprecated. Instead, use yarn.system-metrics-publisher.enabled

17/12/27 11:25:37 INFO mapreduce.JobSubmitter: Submitting tokens for job: job_1514343869308_0001

17/12/27 11:25:38 INFO impl.YarnClientImpl: Submitted application application_1514343869308_0001

17/12/27 11:25:38 INFO mapreduce.Job: The url to track the job:http://sht-sgmhadoopdn-01:8088/proxy/application_1514343869308_0001/

17/12/27 11:25:38 INFO mapreduce.Job: Running job: job_1514343869308_0001

17/12/27 11:25:51 INFO mapreduce.Job: Job job_1514343869308_0001 running in uber mode : false

17/12/27 11:25:51 INFO mapreduce.Job:  map 0% reduce 0%

17/12/27 11:26:14 INFO mapreduce.Job:  map 11% reduce 0%

17/12/27 11:26:15 INFO mapreduce.Job:  map 67% reduce 0%

17/12/27 11:26:29 INFO mapreduce.Job:  map 100% reduce 0%

17/12/27 11:26:32 INFO mapreduce.Job:  map 100% reduce 100%

17/12/27 11:26:34 INFO mapreduce.Job: Job job_1514343869308_0001 completed successfully

17/12/27 11:26:34 INFO mapreduce.Job: Counters: 50

......

[root@sht-sgmhadoopdn-01 hadoop]# hdfs dfs -cat output/*

17/12/27 11:30:08 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable

1    dfsadmin

1    dfs.replication

1    dfs.namenode.secondary.http

1    dfs.namenode.name.dir

1    dfs.datanode.data.dir

也可以通过浏览器访问console，查看详细的分析信息：

ResourceManager -http://172.16.101.58:8088

停止hadoop集群

在 master节点上执行：

[root@sht-sgmhadoopdn-01 hadoop]#stop-yarn.sh

[root@sht-sgmhadoopdn-01 hadoop]#stop-dfs.sh

[root@sht-sgmhadoopdn-01 hadoop]#mr-jobhistory-daemon.sh stop historyserver

参考链接：

http://www.powerxing.com/install-hadoop-cluster/

http://hadoop.apache.org/docs/r2.9.0/hadoop-project-dist/hadoop-common/ClusterSetup.html

另外有需要云服务器可以了解下创新互联scvps.cn，海内外云服务器15元起步，三天无理由+7*72小时售后在线，公司持有idc许可证，提供“云服务器、裸金属服务器、高防服务器、香港服务器、美国服务器、虚拟主机、免备案服务器”等云主机租用服务以及企业上云的综合解决方案，具有“安全稳定、简单易用、服务可用性高、性价比高”等特点与优势，专为企业上云打造定制，能够满足用户丰富、多元化的应用场景需求。