阿里云环境下搭建HadoopHA集群

1. HadoopHA介绍

1.1 hadoop高可用集群的简介

​ hadoop是一个海量数据存储和计算的平台，能够存储PB级以上的数据，并且利用MapReduce可以对这些数据进行并发式计算；hadoop平台采用的是主从式架构（master/slave）即一个master和若干个slave，这个master就是namenode节点，该节点负责存储和管理元数据、配置副本策略、管理数据块的映射信息以及处理客服端的读写请求，由此可见namenode节点的压力还是比较大；众所周知，主从结构的框架虽然容易进行集群资源的管理和调度，但是有个很大的问题，也就是单点故障的问题。假如namenode节点宕机，将造成整个集群瘫痪无法工作；设计者虽然设计了Secondnamenode节点帮助namenode进行资源管理，但这毕竟是冷备份，无法做到实时热备。因此，我们利用zookeeper（集群协调者）来实现hadoop的高可用集群，即一个集群中有两个namenode，当其中一个宕机，另一个能够无缝衔接，做到实时的热备，保证大数据集群的稳定性。

1.2 HadoopHA架构

image-20210811205314030

1.3 自动故障转移工作机制

img

1.4 集群规划

这里使用三台机器搭建集群

hadoop001 hadoop002 hadoop003
NameNode NameNode
JournalNode JournalNode JournalNode
DataNode DataNode DataNode
zk zk zk
ResourceManager ResourceManager
NodeManager NodeManager NodeManager

2. 阿里云服务器的购买

由于我还是学生，资金有限，这里选择简单低配的服务器

2.1 进入阿里云

阿里云-上云就上阿里云 (aliyun.com)

image-20210811210549738

先注册登录

image-20210811210724749

选择右上角的控制台

image-20210811210811101

点击左上角的三条横线

image-20210811210856426

选择ECS云服务器

2.1 选择服务器

点击创建服务器

image-20210811210959212

我这里选择2核8G的抢占式主机，这个比较便宜

image-20210811211433534

注意：上面的区域应该同一个（例如：可用区K，因为不同地区的实例内网不同，而hadoop集群需要使用内网搭建）。我忘记选了，后面随机给我分到一个区了。

image-20210811211656369

然后点击下面的网络配置

2.3 网络选择配置

image-20210811212103986

2.4 系统配置

image-20210811212823155

然后点击分组设置

分组设置保持默认即可，不需要配置

最后确认清晰，然后提交订单即可

image-20210811212953301

然后在控制台的左边栏点击ECS云服务器然后点击示例

image-20210811213140697

可以看出，服务器正在创建

2.5 配置服务器远程连接

等待服务器创建完成后，我们需要使用远程连接工具来连接服务器

这里我使用的是xshell进行连接

image-20210811213408970

这里我们可以看到，每个服务器有两个IP地址，一个公网ip，一个是内网ip；我们搭建hadoop集群是使用内网ip

首先检查安全组对外的22号端口是否开放

image-20210811213806183

image-20210811213907165

注意：我这里只有一个安全组，因此所有实例都配置在这个安全组里面。

image-20210811214040116

进去后检查22号端口是否开放，若没够开放，需要自己手动添加一个。

使用公网IP连接ECS

image-20210811214218888

复制这里的公网IP，然后根据自己设置的用户名和密码使用Xshell连接即可。

image-20210811214616083

这里我三台机器已经连接

3. 配置网络

3.1 配置主机名

这里由于主机名已经预设好了，因此不用配置了

3.2 配置hosts文件

[root@hadoop001 ~]# vim /etc/hosts

image-20210811215625109

需要在三台机器上都配置

3.2 检查防火墙

一般企业内部的集群之间会关闭防火墙，防止集群间通信被阻挡；企业会在外部网口处统一设置防火墙

image-20210811220101988

阿里云默认的机器防火墙是关闭的，因此可以不用管了

3.3 配置SSH

首先执行：ssh-keygen -t rsa指令

[root@hadoop001 ~]# ssh-keygen -t rsa
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:
SHA256:Q7rbZb6Eu2aHpYEhZtHw97bCuIo1Tl14DALrKwiHb9U root@hadoop001
The key's randomart image is:
+---[RSA 2048]----+
|  . .o           |
|   o...          |
|  . ..o o        |
| o  +o.B .       |
|o oo..EoS o      |
|oo o ..*.+..     |
|o + + + +==      |
| o = . +=B.      |
|  . o.oo+oo.     |
+----[SHA256]-----+

连续按下三次回车，生成公钥和私钥

将公钥分发到其他集群

image-20210811220341815

以此类推，每台机器都要配置。

然后我们使用ssh 主机名 测试能够互相登录

image-20210811220726236

4. 安装JDK

由于hadoop时使用java语言编写的，因此我们需要java的运行环境

4.1 先卸载预装的jdk

image-20210811221413417

经过搜索，没有预装jdk，因此我们不用卸载了

4.2 上传软件安装包

image-20210811221508227

这里将三个安装包都上传

4.3 解压到指定目录

[root@hadoop001 software]# tar -zxvf jdk-8u231-linux-x64.tar.gz -C /opt/module/

4.4 配置环境变量

[root@hadoop001 jdk1.8.0_231]# pwd
/opt/module/jdk1.8.0_231
[root@hadoop001 jdk1.8.0_231]# vim /etc/profile
# 在profile文件的末尾添加：

export JAVA_HOME=/opt/module/jdk1.8.0_231
export PATH=$PATH:$JAVA_HOME/bin

# 保存退出后，使用source指令使得配置生效
[root@hadoop001 jdk1.8.0_231]# source /etc/profile

# 执行java -version指令
[root@hadoop001 ~]# java -version
java version "1.8.0_231"
Java(TM) SE Runtime Environment (build 1.8.0_231-b11)
Java HotSpot(TM) 64-Bit Server VM (build 25.231-b11, mixed mode)
[root@hadoop001 ~]# 
# 可以看出已经配置成功

其他两个节点也需要配置

5.搭建zookeeper集群

注意：我所有的软件都安装在/opt/module下面，可以按照个人习惯修改

5.1 解压zookeeper安装包

[root@hadoop001 software]# tar -zxvf zookeeper-3.4.10.tar.gz -C /opt/module/

[root@hadoop001 module]# ll
total 8
drwxr-xr-x  7   10  143 4096 Oct  5  2019 jdk1.8.0_231
drwxr-xr-x 10 1001 1001 4096 Mar 23  2017 zookeeper-3.4.10

5.2 配置zookeeper

（1）、在/opt/module/zookeeper-3.4.10下创建zkData目录

[root@hadoop001 zookeeper-3.4.10]# mkdir zkData

（2）、将/opt/module/zookeeper-3.4.10/conf这个目录下的zoo_sample.cfg修改为zoo.cfg

[root@hadoop001 conf]# ll
total 12
-rw-rw-r-- 1 1001 1001  535 Mar 23  2017 configuration.xsl
-rw-rw-r-- 1 1001 1001 2161 Mar 23  2017 log4j.properties
-rw-rw-r-- 1 1001 1001  922 Mar 23  2017 zoo_sample.cfg
[root@hadoop001 conf]# mv zoo_sample.cfg zoo.cfg
[root@hadoop001 conf]#

（3）、配置zoo.cfg

# 修改dataDir
dataDir=/opt/module/zookeeper-3.4.10/zkData

# 添加集群配置
server.1=hadoop001:2888:3888
server.2=hadoop002:2888:3888
server.3=hadoop003:2888:3888

（4）、在zkData中创建myid文件

[root@hadoop001 conf]# cd /opt/module/zookeeper-3.4.10/zkData/
# 进入zkData下面创建
[root@hadoop001 zkData]# touch myid
[root@hadoop001 zkData]# vim myid 

# myid中添加服务器编号，也就上面zoo.cfg中server.后面的数字

image-20210811224208585

（5）、将zookeeper分发到其他两个节点

[root@hadoop001 module]# scp -r zookeeper-3.4.10/ hadoop002:`pwd`

（6）、修改各个节点对应的myid中的值

image-20210811224505577

5.3 启动zookeeper

我们需要在三台节点上执行：

[root@hadoop001 zookeeper-3.4.10]# bin/zkServer.sh start

image-20210811224747365

各个节点都出现QuorumPeerMain进程，并也已经选举产生了leader，因此zookeeper集群安装完成。

6. 搭建hadoopHA

6.1 解压Hadoop

[root@hadoop001 software]# tar -zxvf hadoop-2.7.2.tar.gz -C /opt/module/

6.2 配置环境变量

[root@hadoop001 software]# cd /opt/module/hadoop-2.7.2/
[root@hadoop001 hadoop-2.7.2]# pwd
/opt/module/hadoop-2.7.2
[root@hadoop001 hadoop-2.7.2]# vim /etc/profile

# 在profile文件最后添加
export HADOOP_HOME=/opt/module/hadoop-2.7.2
export PATH=$PATH:$HADOOP_HOME/bin:$HADOOP_HOME/sbin

# source使文件生效
[root@hadoop001 hadoop-2.7.2]# source /etc/profile

[root@hadoop001 hadoop-2.7.2]# hadoop version
Hadoop 2.7.2
Subversion Unknown -r Unknown
Compiled by root on 2017-05-22T10:49Z
Compiled with protoc 2.5.0
From source with checksum d0fda26633fa762bff87ec759ebe689c
This command was run using /opt/module/hadoop-2.7.2/share/hadoop/common/hadoop-common-2.7.2.jar
[root@hadoop001 hadoop-2.7.2]#

# 可以看出，配置完成

6.3 修改配置文件

# 进入配置文件目录
[root@hadoop001 hadoop-2.7.2]# cd etc/hadoop/
[root@hadoop001 hadoop]# pwd
/opt/module/hadoop-2.7.2/etc/hadoop

(1)、编辑hadoop-env.sh文件

[root@hadoop001 hadoop]# vim hadoop-env.sh

# 修改JAVA_HOME配置
export JAVA_HOME=/opt/module/jdk1.8.0_231

# 修改HADOOP_CONF_DIR配置
export HADOOP_CONF_DIR=/opt/module/hadoop-2.7.2/etc/hadoop

(2)、编写slaves文件

[root@hadoop001 hadoop]# vim slaves 

# 在slaves中添加
hadoop001
hadoop002
hadoop003

image-20210811230535340

(3)、配置core-site.xml文件

# 在configuration标签中添加
	    <!-- 指定在Zookeeper上注册的节点的名字 -->
        <property>
            <name>fs.defaultFS</name>
            <value>hdfs://ns</value>
        </property>
        <!-- 指定Hadoop数据存放目录 -->
        <property>
            <name>hadoop.tmp.dir</name>
            <value>/opt/module/hadoop-2.7.2/data</value>
        </property>
        <!-- 指定zookeeper的连接地址 -->
        <property>
            <name>ha.zookeeper.quorum</name>
            <value>hadoop001:2181,hadoop002:2181,hadoop003:2181</value>
        </property>

(4)、编辑hdfs-site.xml文件

# 在configuration标签中添加
	<!-- 绑定在Zookeeper上注册的节点名 -->
	<property>
	    <name>dfs.nameservices</name>
	    <value>ns</value>
	</property>
	<!-- ns集群下有两个namenode，分别为nn1, nn2 -->
	<property>
	    <name>dfs.ha.namenodes.ns</name>
	    <value>nn1,nn2</value>
	</property>
	<!--nn1的RPC通信-->
	<property>
	    <name>dfs.namenode.rpc-address.ns.nn1</name>
	    <value>hadoop001:9000</value>
	</property>
	<!--nn1的http通信-->
	<property>
	    <name>dfs.namenode.http-address.ns.nn1</name>
	    <value>hadoop001:50070</value>
	</property>
	<!-- nn2的RPC通信地址 -->
	<property>
	    <name>dfs.namenode.rpc-address.ns.nn2</name>
	    <value>hadoop002:9000</value>
	</property>
	<!-- nn2的http通信地址 -->
	<property>
	    <name>dfs.namenode.http-address.ns.nn2</name>
	    <value>hadoop002:50070</value>
	</property>
	<!--指定namenode的元数据在JournalNode上存放的位置，这样，namenode2可以从journalnode集群里的指定位置上获取信息，达到热备效果-->
	<property>
	    <name>dfs.namenode.shared.edits.dir</name>
	    <value>qjournal://hadoop001:8485;hadoop002:8485;hadoop003:8485/ns</value>
	</property>
	<!-- 指定JournalNode在本地磁盘存放数据的位置 -->
	<property>
	    <name>dfs.journalnode.edits.dir</name>
	    <value>/opt/module/hadoop-2.7.2/data/journal</value>
	</property>
	<!-- 开启NameNode故障时自动切换 -->
	<property>
	    <name>dfs.ha.automatic-failover.enabled</name>
	    <value>true</value>
	</property>
	<!-- 配置失败自动切换实现方式 -->
	<property>
	    <name>dfs.client.failover.proxy.provider.ns</name>
	    <value>org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider</value>
	</property>
	<!-- 配置隔离机制 -->
	<property>
	    <name>dfs.ha.fencing.methods</name>
	    <value>sshfence</value>
	</property>
	<!-- 使用隔离机制时需要ssh免登陆 -->
	<property>
	    <name>dfs.ha.fencing.ssh.private-key-files</name>
	    <value>/root/.ssh/id_rsa</value>
	</property>
	<!--配置namenode存放元数据的目录，可以不配置，如果不配置则默认放到hadoop.tmp.dir下-->
	<property>    
	    <name>dfs.namenode.name.dir</name>    
	    <value>file:///opt/module/hadoop-2.7.2/data/hdfs/name</value>    
	</property>    
	<!--配置datanode存放元数据的目录，可以不配置，如果不配置则默认放到hadoop.tmp.dir下-->
	<property>    
	    <name>dfs.datanode.data.dir</name>    
	    <value>file:///opt/module/hadoop-2.7.2/data/hdfs/data</value>    
	</property>
	<!--配置副本数量-->    
	<property>    
	    <name>dfs.replication</name>    
	    <value>3</value>    
	</property>   
	<!--设置用户的操作权限，false表示关闭权限验证，任何用户都可以操作-->               
	<property>    
	    <name>dfs.permissions</name>    
	    <value>false</value>    
	</property>

(5)、编辑mapred-site.xml文件

[root@hadoop001 hadoop]# mv mapred-site.xml.template mapred-site.xml
[root@hadoop001 hadoop]# vim mapred-site.xml

# 在configuration标签中添加
	<property>    
	    <name>mapreduce.framework.name</name>    
	    <value>yarn</value>    
	</property>

(6)、编辑yarn-site.xml文件

# 在configuration标签中添加
	<!--配置yarn的高可用-->
	<property>
	    <name>yarn.resourcemanager.ha.enabled</name>
	    <value>true</value>
	</property>
	<!--指定两个resourcemaneger的名称-->
	<property>
	    <name>yarn.resourcemanager.ha.rm-ids</name>
	    <value>rm1,rm2</value>
	</property>
	<!--配置rm1的主机-->
	<property>
	    <name>yarn.resourcemanager.hostname.rm1</name>
	    <value>hadoop001</value>
	</property>
	<!--配置rm2的主机-->
	<property>
	    <name>yarn.resourcemanager.hostname.rm2</name>
	    <value>hadoop003</value>
	</property>
	<!--开启yarn恢复机制-->
	<property>
	    <name>yarn.resourcemanager.recovery.enabled</name>
	    <value>true</value>
	</property>
	<!--执行rm恢复机制实现类-->
	<property>
	    <name>yarn.resourcemanager.store.class</name>
	    <value>org.apache.hadoop.yarn.server.resourcemanager.recovery.ZKRMStateStore</value>
	</property>
	<!--配置zookeeper的地址-->
	<property>
	    <name>yarn.resourcemanager.zk-address</name>
	    <value>hadoop001:2181,hadoop002:2181,hadoop003:2181</value>
	</property>
	<!--执行yarn集群的别名-->
	<property>
	    <name>yarn.resourcemanager.cluster-id</name>
	    <value>ns-yarn</value>
	</property>
	<!-- 指定nodemanager启动时加载server的方式为shuffle server -->
	<property>    
	    <name>yarn.nodemanager.aux-services</name>    
	    <value>mapreduce_shuffle</value>    
	</property>  
	<!-- 指定resourcemanager地址 -->
	<property>
	    <name>yarn.resourcemanager.hostname</name>
	    <value>hadoop003</value>
	</property>

6.4 分发hadoop

# 分别拷贝到hadoop002和hadoop003
[root@hadoop001 module]# scp -r hadoop-2.7.2/ hadoop002:`pwd`

[root@hadoop001 module]# scp -r hadoop-2.7.2/ hadoop003:`pwd`

# 然后在其他两个机器上配置环境变量

6.5 启动hadoopHA集群

注意： zookeeper需要先开启

(1)、在hadoop001上格式化Zookeeper

[root@hadoop001 ~]# hdfs zkfc -formatZK

image-20210811233100823

出现上面这种，说明成功

(2)、在三台节点上启动journalNode

[root@hadoop001 ~]# hadoop-daemon.sh start journalnode
starting journalnode, logging to /opt/module/hadoop-2.7.2/logs/hadoop-root-journalnode-hadoop001.out
[root@hadoop001 ~]# jps
4945 Jps
2418 QuorumPeerMain
4890 JournalNode
[root@hadoop001 ~]#

注意： 三台机器都要启动

image-20210811233743879

(3)、在hadoop001上执行格式化指令并启动namenode

[root@hadoop001 ~]# hadoop namenode -format

image-20210811233858702

执行成功

启动namenode

[root@hadoop001 ~]# hadoop-daemon.sh start namenode
starting namenode, logging to /opt/module/hadoop-2.7.2/logs/hadoop-root-namenode-hadoop001.out
[root@hadoop001 ~]# jps
2418 QuorumPeerMain
4890 JournalNode
5485 Jps
5407 NameNode
[root@hadoop001 ~]#

image-20210811234049028

(4)、在hadoop002上执行格式化指令并启动namenode

[root@hadoop002 ~]# hadoop namenode -bootstrapStandBy

image-20210811234258163

格式化成功

启动namenode：

[root@hadoop002 ~]# hadoop-daemon.sh start namenode
starting namenode, logging to /opt/module/hadoop-2.7.2/logs/hadoop-root-namenode-hadoop002.out
[root@hadoop002 ~]# jps
2256 Jps
1830 QuorumPeerMain
2182 NameNode
2025 JournalNode
[root@hadoop002 ~]#

(5)、启动三台节点的datanode

# 三台机器都要执行
[root@hadoop001 ~]# hadoop-daemon.sh start datanode

image-20210811234533814

(6)、在hadoop001和hadoop002上启动FailOverController

[root@hadoop001 ~]# hadoop-daemon.sh start zkfc
starting zkfc, logging to /opt/module/hadoop-2.7.2/logs/hadoop-root-zkfc-hadoop001.out
[root@hadoop001 ~]# jps
2418 QuorumPeerMain
5990 DFSZKFailoverController
5752 DataNode
4890 JournalNode
6074 Jps
5407 NameNode
[root@hadoop001 ~]#

(7)、在hadoop003上执行start-yarn.sh

[root@hadoop003 ~]# start-yarn.sh 
starting yarn daemons
starting resourcemanager, logging to /opt/module/hadoop-2.7.2/logs/yarn-root-resourcemanager-hadoop003.out
hadoop002: starting nodemanager, logging to /opt/module/hadoop-2.7.2/logs/yarn-root-nodemanager-hadoop002.out
hadoop001: starting nodemanager, logging to /opt/module/hadoop-2.7.2/logs/yarn-root-nodemanager-hadoop001.out
hadoop003: starting nodemanager, logging to /opt/module/hadoop-2.7.2/logs/yarn-root-nodemanager-hadoop003.out
[root@hadoop003 ~]# jps
1856 QuorumPeerMain
2609 Jps
2214 ResourceManager
2007 JournalNode
2089 DataNode
2319 NodeManager
[root@hadoop003 ~]#

(8)、在hadoop001上执行yarn-daemon.sh start resourcemanager

[root@hadoop001 ~]# yarn-daemon.sh start resourcemanager

(9)、最终节点状态

hadoop001: (8个进程)

image-20210811235033248

hadoop002: (7个进程)

image-20210811235109001

hadoop003: (6个进程)

image-20210811235134334

6.6 hadoopweb ui界面

这里需要先去阿里云的安全组开放50070端口，然后才能访问。

image-20210811235504261

image-20210811235523089

image-20210811235610252

上面可以看出，各个组件的webui均能访问。

6.7 测试hdfs集群

我们上传一些文件：

# 使用put指令上传hadoop安装包
[root@hadoop001 software]# hadoop fs -put hadoop-2.7.2.tar.gz /
[root@hadoop001 software]# hadoop fs -ls /
Found 1 items
-rw-r--r--   3 root supergroup  197657687 2021-08-11 23:57 /hadoop-2.7.2.tar.gz
[root@hadoop001 software]#

webui查看

image-20210811235836226

上传成功

6.8 测试yarn集群

运行一个mapreduce任务：

# 计算圆周率
[root@hadoop001 hadoop-2.7.2]# hadoop jar share/hadoop/mapreduce/hadoop-mapreduce-examples-2.7.2.jar pi 20 20
Number of Maps  = 20
Samples per Map = 20
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
Wrote input for Map #10
Wrote input for Map #11
Wrote input for Map #12
Wrote input for Map #13
Wrote input for Map #14
Wrote input for Map #15
Wrote input for Map #16
Wrote input for Map #17
Wrote input for Map #18
Wrote input for Map #19
Starting Job
21/08/12 00:00:47 INFO client.ConfiguredRMFailoverProxyProvider: Failing over to rm2
21/08/12 00:00:48 INFO input.FileInputFormat: Total input paths to process : 20
21/08/12 00:00:48 INFO mapreduce.JobSubmitter: number of splits:20
21/08/12 00:00:48 INFO mapreduce.JobSubmitter: Submitting tokens for job: job_1628696898160_0001
21/08/12 00:00:48 INFO impl.YarnClientImpl: Submitted application application_1628696898160_0001
21/08/12 00:00:48 INFO mapreduce.Job: The url to track the job: http://hadoop003:8088/proxy/application_1628696898160_0001/
21/08/12 00:00:48 INFO mapreduce.Job: Running job: job_1628696898160_0001
21/08/12 00:00:55 INFO mapreduce.Job: Job job_1628696898160_0001 running in uber mode : false
21/08/12 00:00:55 INFO mapreduce.Job:  map 0% reduce 0%
21/08/12 00:01:07 INFO mapreduce.Job:  map 30% reduce 0%
21/08/12 00:01:09 INFO mapreduce.Job:  map 45% reduce 0%
21/08/12 00:01:10 INFO mapreduce.Job:  map 65% reduce 0%
21/08/12 00:01:11 INFO mapreduce.Job:  map 100% reduce 0%
21/08/12 00:01:12 INFO mapreduce.Job:  map 100% reduce 100%
21/08/12 00:01:13 INFO mapreduce.Job: Job job_1628696898160_0001 completed successfully
21/08/12 00:01:13 INFO mapreduce.Job: Counters: 49
	File System Counters
		FILE: Number of bytes read=446
		FILE: Number of bytes written=2524447
		FILE: Number of read operations=0
		FILE: Number of large read operations=0
		FILE: Number of write operations=0
		HDFS: Number of bytes read=5030
		HDFS: Number of bytes written=215
		HDFS: Number of read operations=83
		HDFS: Number of large read operations=0
		HDFS: Number of write operations=3
	Job Counters 
		Launched map tasks=20
		Launched reduce tasks=1
		Data-local map tasks=20
		Total time spent by all maps in occupied slots (ms)=253189
		Total time spent by all reduces in occupied slots (ms)=2746
		Total time spent by all map tasks (ms)=253189
		Total time spent by all reduce tasks (ms)=2746
		Total vcore-milliseconds taken by all map tasks=253189
		Total vcore-milliseconds taken by all reduce tasks=2746
		Total megabyte-milliseconds taken by all map tasks=259265536
		Total megabyte-milliseconds taken by all reduce tasks=2811904
	Map-Reduce Framework
		Map input records=20
		Map output records=40
		Map output bytes=360
		Map output materialized bytes=560
		Input split bytes=2670
		Combine input records=0
		Combine output records=0
		Reduce input groups=2
		Reduce shuffle bytes=560
		Reduce input records=40
		Reduce output records=0
		Spilled Records=80
		Shuffled Maps =20
		Failed Shuffles=0
		Merged Map outputs=20
		GC time elapsed (ms)=6425
		CPU time spent (ms)=6740
		Physical memory (bytes) snapshot=5434896384
		Virtual memory (bytes) snapshot=44580233216
		Total committed heap usage (bytes)=4205838336
	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=2360
	File Output Format Counters 
		Bytes Written=97
Job Finished in 26.222 seconds
Estimated value of Pi is 3.17000000000000000000

这里计算成功。

6.9 关闭集群

在hadoop001上执行stop-all.sh
[root@hadoop001 ~]# stop-all.sh 
This script is Deprecated. Instead use stop-dfs.sh and stop-yarn.sh
Stopping namenodes on [hadoop001 hadoop002]
hadoop002: stopping namenode
hadoop001: stopping namenode
hadoop002: stopping datanode
hadoop001: stopping datanode
hadoop003: stopping datanode
Stopping journal nodes [hadoop001 hadoop002 hadoop003]
hadoop003: stopping journalnode
hadoop001: stopping journalnode
hadoop002: stopping journalnode
Stopping ZK Failover Controllers on NN hosts [hadoop001 hadoop002]
hadoop002: stopping zkfc
hadoop001: stopping zkfc
stopping yarn daemons
stopping resourcemanager
hadoop003: stopping nodemanager
hadoop002: stopping nodemanager
hadoop001: stopping nodemanager
hadoop003: nodemanager did not stop gracefully after 5 seconds: killing with kill -9
hadoop002: nodemanager did not stop gracefully after 5 seconds: killing with kill -9
hadoop001: nodemanager did not stop gracefully after 5 seconds: killing with kill -9
no proxyserver to stop

已经关闭了集群。

7. 注意点

1、 集群第一次启动由于需要初始化，因此启动比较麻烦。

第一次启动成功后，下次启动，在hadoop001上执行start-all.sh就可以了。

[root@hadoop001 ~]# start-all.sh

2、启动hadoopHA集群前一定要先启动zookeeper集群

3: stopping datanode
Stopping journal nodes [hadoop001 hadoop002 hadoop003]
hadoop003: stopping journalnode
hadoop001: stopping journalnode
hadoop002: stopping journalnode
Stopping ZK Failover Controllers on NN hosts [hadoop001 hadoop002]
hadoop002: stopping zkfc
hadoop001: stopping zkfc
stopping yarn daemons
stopping resourcemanager
hadoop003: stopping nodemanager
hadoop002: stopping nodemanager
hadoop001: stopping nodemanager
hadoop003: nodemanager did not stop gracefully after 5 seconds: killing with kill -9
hadoop002: nodemanager did not stop gracefully after 5 seconds: killing with kill -9
hadoop001: nodemanager did not stop gracefully after 5 seconds: killing with kill -9
no proxyserver to stop


已经关闭了集群。

## 7. 注意点

1、 集群第一次启动由于需要初始化，因此启动比较麻烦。

第一次启动成功后，下次启动，在hadoop001上执行start-all.sh就可以了。

```shell
[root@hadoop001 ~]# start-all.sh

2、启动hadoopHA集群前一定要先启动zookeeper集群

# hadoop
Logo
技术共进，成长同行——讯飞AI开发者社区

技术共进，成长同行——讯飞AI开发者社区

更多推荐

论文笔记：AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models（AlphaEdit）

论文发表于人工智能顶会ICLR（基于定位和修改的模型编辑方法（针对和等）会破坏LLM中最初保存的知识，特别是在顺序编辑场景。为此，本文提出AlphaEdit：1、在将保留知识应用于参数之前，将扰动投影到保留知识的零空间上。2、从理论上证明，这种预测确保了在查询保留的知识时，编辑后的LLM的输出保持不变，从而减轻中断问题。3、对各种LLM（包括LLaMA3、GPT2XL和GPT-J）的广泛实验表明，

avatar 讯飞AI开发者社区

2025 年就业竞争激烈！当1222 万毕业生遭遇 “35 岁门槛” 与 AI 革命的双重冲击

人工智能

avatar 讯飞AI开发者社区

论文解读：MASS-EDITING MEMORY IN A TRANSFORMER（MEMIT）

论文发表于人工智能顶会ICLR（在模型编辑方法中，主要局限于更新单个事实。因此，基于ROME，本文开发了MEMIT，在大模型GPT-J（6B）和GPT-NeoX（20B）上实现了数千的批量编辑。阅读本文请同时参考原始论文图表。

avatar 讯飞AI开发者社区