hadoop动态添加删除节点datanode及恢复

1. 配置系统环境

主机名,ssh互信,环境变量等

本文略去jdk安装，请将datanode的jdk安装路径与/etc/hadoop/hadoop-evn.sh中的java_home保持一致，版本hadoop2.7.5

修改/etc/sysconfig/network

然后执行命令
hostname 主机名
这个时候可以注销一下系统，再重登录之后就行了

[root@localhost ~]# hostname
localhost.localdomain
[root@localhost ~]# hostname -i
:: 127.0.0.1
[root@localhost ~]#
[root@localhost ~]# cat /etc/sysconfig/network
# Created by anaconda
NETWORKING=yes
HOSTNAME=slave2
GATEWAY=192.168.48.2
# oracle-rdbms-server-11gR2-preinstall : Add NOZEROCONF=yes
NOZEROCONF=yes
[root@localhost ~]# hostname slave2
[root@localhost ~]# hostname
slave2
[root@localhost ~]# su - hadoop
Last login: Sat Feb  :: CST  on pts/
[hadoop@slave2 ~]$ su - root

建datanode目录并改所有者

（此处的具体路径值，请参照namenode中/usr/hadoop/hadoop-2.7.5/etc/hadoop/hdfs-site.xml,core-site.xml中的dfs.name.dir，dfs.data.dir,dfs.tmp.dir等）

Su - root

# mkdir -p /usr/local/hadoop-2.7.5/tmp/dfs/data

# chmod -R 777 /usr/local/hadoop-2.7.5/tmp

# chown -R hadoop:hadoop /usr/local/hadoop-2.7.5

[root@slave2 ~]# mkdir -p /usr/local/hadoop-2.7./tmp/dfs/data
[root@slave2 ~]# chmod -R  /usr/local/hadoop-2.7./tmp
 [root@slave2 ~]# chown -R hadoop:hadoop /usr/local/hadoop-2.7.
 [root@slave2 ~]# pwd
/root
[root@slave2 ~]# cd /usr/local/
[root@slave2 local]# ll
total
drwxr-xr-x.  root   root    Mar    bin
drwxr-xr-x.  root   root     Jun    etc
drwxr-xr-x.  root   root     Jun    games
drwxr-xr-x   hadoop hadoop  Feb  : hadoop-2.7.
drwxr-xr-x.  root   root     Jun    include
drwxr-xr-x.  root   root     Jun    lib
drwxr-xr-x.  root   root     Jun    lib64
drwxr-xr-x.  root   root     Jun    libexec
drwxr-xr-x.  root   root     Jun    sbin
drwxr-xr-x.  root   root    Dec    share
drwxr-xr-x.  root   root     Jun    src
[root@slave2 local]#

ssh互信，即实现 master-->slave2免密码

master:

[root@hadoop-master ~]# cat /etc/hosts
127.0.0.1   localhost localhost.localdomain localhost4 localhost4.localdomain4
::1         localhost localhost.localdomain localhost6 localhost6.localdomain6
192.168.48.129    hadoop-master
192.168.48.132    slave1
192.168.48.131    slave2
[hadoop@hadoop-master ~]$ scp /usr/hadoop/.ssh/authorized_keys hadoop@slave2:/usr/hadoop/.ssh
The authenticity of host 'slave2 (192.168.48.131)' can't be established.
ECDSA key fingerprint is 1e:cd:d1:3d:b0:5b:62:45:a3:63:df:c7:7a:0f:b8:7c.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added 'slave2,192.168.48.131' (ECDSA) to the list of known hosts.
hadoop@slave2's password:
authorized_keys
[hadoop@hadoop-master ~]$ ssh hadoop@slave2
Last login: Sat Feb 24 18:27:33 2018
[hadoop@slave2 ~]$
[hadoop@slave2 ~]$ exit
logout
Connection to slave2 closed.
[hadoop@hadoop-master ~]$

2. 修改namenode节点的slave文件,增加新节点信息

[hadoop@hadoop-master hadoop]$ pwd
/usr/hadoop/hadoop-2.7.5/etc/hadoop
[hadoop@hadoop-master hadoop]$ vi slaves
slave1
slave2

3. 在namenode节点上,将hadoop-2.7.3复制到新节点上,并在新节点上删除data和logs目录中的文件

Master

[hadoop@hadoop-master ~]$ scp -R hadoop-2.7.5 hadoop@slave2:/usr/hadoop

Slave2
[hadoop@slave2 hadoop-2.7.5]$ ll
total 124
drwxr-xr-x 2 hadoop hadoop  4096 Feb 24 14:29 bin
drwxr-xr-x 3 hadoop hadoop    19 Feb 24 14:30 etc
drwxr-xr-x 2 hadoop hadoop   101 Feb 24 14:30 include
drwxr-xr-x 3 hadoop hadoop    19 Feb 24 14:29 lib
drwxr-xr-x 2 hadoop hadoop  4096 Feb 24 14:29 libexec
-rw-r--r-- 1 hadoop hadoop 86424 Feb 24 18:44 LICENSE.txt
drwxrwxr-x 2 hadoop hadoop  4096 Feb 24 14:30 logs
-rw-r--r-- 1 hadoop hadoop 14978 Feb 24 18:44 NOTICE.txt
-rw-r--r-- 1 hadoop hadoop  1366 Feb 24 18:44 README.txt
drwxr-xr-x 2 hadoop hadoop  4096 Feb 24 14:29 sbin
drwxr-xr-x 4 hadoop hadoop    29 Feb 24 14:30 share
[hadoop@slave2 hadoop-2.7.5]$ pwd
/usr/hadoop/hadoop-2.7.5
[hadoop@slave2 hadoop-2.7.5]$ rm -R logs/*

4. 启动新datanode的datanode和nodemanger进程

先确认namenode和当前的datanode中，etc/hoadoop/excludes文件中无待加入的主机，再进行下面操作

[hadoop@slave2 hadoop-2.7.]$ sbin/hadoop-daemon.sh start datanode
starting datanode, logging to /usr/hadoop/hadoop-2.7./logs/hadoop-hadoop-datanode-slave2.out
[hadoop@slave2 hadoop-2.7.]$ sbin/yarn-daemon.sh start nodemanager
starting datanode, logging to /usr/hadoop/hadoop-2.7./logs/yarn-hadoop-datanode-slave2.out
[hadoop@slave2 hadoop-2.7.]$
[hadoop@slave2 hadoop-2.7.]$ jps
 DataNode
 NodeManager
 Jps
[hadoop@slave2 ~]$

5、在NameNode上刷新节点

[hadoop@hadoop-master ~]$ hdfs dfsadmin -refreshNodes
Refresh nodes successful
[hadoop@hadoop-master ~]$sbin/start-balancer.sh

6. 在namenode查看当前集群情况,

确认节点已经正常加入

[hadoop@hadoop-master hadoop]$ hdfs dfsadmin -report
Configured Capacity:  (54.63 GB)
Present Capacity:  (14.42 GB)
DFS Remaining:  (14.42 GB)
DFS Used:  ( KB)
DFS Used%: 0.00%
Under replicated blocks:
Blocks with corrupt replicas:
Missing blocks:
Missing blocks (with replication factor ): 

-------------------------------------------------
Live datanodes (2):

Name: 192.168.48.131:50010 (slave2)
Hostname: 183.221.250.11
Decommission Status : Normal
Configured Capacity: 38588669952 (35.94 GB)
DFS Used: 8192 (8 KB)
Non DFS Used: 36887191552 (34.35 GB)
DFS Remaining: 1701470208 (1.58 GB)
DFS Used%: 0.00%
DFS Remaining%: 4.41%
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: Thu Mar 01 19:36:33 PST 2018

Name: 192.168.48.132: (slave1)
Hostname: slave1
Decommission Status : Normal
Configured Capacity:  (18.70 GB)
DFS Used:  ( KB)
Non DFS Used:  (5.86 GB)
DFS Remaining:  (12.84 GB)
DFS Used%: 0.00%
DFS Remaining%: 68.67%
Configured Cache Capacity:  ( B)
Cache Used:  ( B)
Cache Remaining:  ( B)
Cache Used%: 100.00%
Cache Remaining%: 0.00%
Xceivers:
Last contact: Thu Mar  :: PST 

[hadoop@hadoop-master hadoop]$

7动态删除datanode

7.1配置NameNode的hdfs-site.xml，

适当减小dfs.replication副本数，增加dfs.hosts.exclude配置

[hadoop@hadoop-master hadoop]$ pwd
/usr/hadoop/hadoop-2.7./etc/hadoop
[hadoop@hadoop-master hadoop]$ cat hdfs-site.xml
<configuration>
<property>
      <name>dfs.replication</name>
      <value></value>
</property>
  <property>
      <name>dfs.name.dir</name>
      <value>/usr/local/hadoop-2.7./tmp/dfs/name</value>
</property>
    <property>
      <name>dfs.data.dir</name>
      <value>/usr/local/hadoop-2.7./tmp/dfs/data</value>
    </property>
<property>
    <name>dfs.hosts.exclude</name>
    <value>/usr/hadoop/hadoop-2.7.5/etc/hadoop/excludes</value>
  </property>

</configuration>

7.2在namenode对应路径（/etc/hadoop/）下新建excludes文件，

并写入待删除DataNode的ip或域名

[hadoop@hadoop-master hadoop]$ pwd
/usr/hadoop/hadoop-2.7.5/etc/hadoop
[hadoop@hadoop-master hadoop]$ vi excludes
####slave2
192.168.48.131[hadoop@hadoop-master hadoop]$ 

7.3在NameNode上刷新所有DataNode

hdfs dfsadmin -refreshNodes
sbin/start-balancer.sh

7.4在namenode查看当前集群情况,

确认信节点已经正常删除,结果中已无slave2

[hadoop@hadoop-master hadoop]$ hdfs dfsadmin -report 

或者可以在web检测界面（ip:50070）上可以观测到DataNode逐渐变为Dead。

http://192.168.48.129:50070/

在datanode项,Admin state已经由“In Service“变为”Decommissioned“，则表示删除成功

7.5停止已删除的节点相关进程

[hadoop@slave2 hadoop-2.7.]$ jps
 Jps
 DataNode
 NodeManager
[hadoop@slave2 hadoop-2.7.]$ sbin/hadoop-daemon.sh stop datanode
stopping datanode
[hadoop@slave2 hadoop-2.7.]$ sbin/yarn-daemon.sh stop nodemanager
stopping nodemanager
[hadoop@slave2 hadoop-2.7.]$ jps
 Jps
[hadoop@slave2 hadoop-2.7.]$ 

8恢复已删除节点

执行7.2 中删除相关信息，然后4，5，6即可