Hadoop添加datanode或者tasktracker节点

1 规划新添加的hadoop节点的角色和配置

1.1 角色

slave: 即datanode节点或者tasktracker节点

1.2 配置

主机名：hadoop03

IP: 192.168.88.173

hadoop用户：xiaoyu

2. 部署新的节点

2.1 安装操作系统

2.3 关闭不必要的服务

建议只保留以下服务：abrt-ccpp abrt-oops autofs crond haldaemon lvm2-monitor mdmonitor messagebus netfs network nfslock ntpd portreserve rsyslog sshd udev-post

2.4 网络配置

2.4.1 网卡地址

示例配置文件如下，请根据实际所处的网络配置。

# cat /etc/sysconfig/network-scripts/ifcfg-eth0

DEVICE=eth0

TYPE=Ethernet

ONBOOT=yes

NM_CONTROLLED=no

BOOTPROTO=none

IPADDR= 192.168.88.173

NETMASK= 255.255.255.0

GATEWAY= 192.168.88.2

DNS1= 192.168.88.2

IPV6INIT=no

USERCTL=no

2.4.2 修改主机名

# sudo vim /etc/sysconfig/network

修改HOSTNAME值为主机名

HOSTNAME=hadoop03

# hostname hadoop03

2.4.3 配置密钥对

[xiaoyu@hadoop03 ~] $ ssh-keygen -t rsa
Generating public/private rsa key pair.
Enter file in which to save the key (/home/xiaoyu/.ssh/id_rsa):
Enter passphrase (empty for no passphrase):
Enter same passphrase again:
Your identification has been saved in /home/xiaoyu/.ssh/id_rsa.
Your public key has been saved in /home/xiaoyu/.ssh/id_rsa.pub.
The key fingerprint is:
45:41:8d:17:3b:0c:20:e0:5d:3f:38:ed:1f:e6:b9:7a xiaoyu@hadoop03
The key's randomart image is:
+--[ RSA 2048]----+
| ... oo=+.. |
| . . o =.oo. |
| . . o =.+ |
| + . . |
| S . o |
| + o |
| + |
| E. |
| .o. |
+-----------------+
[xiaoyu@hadoop03 ~]$

可以把这个操作写成一行命令，甚至写到脚本中：

$ expect -c "spawn ssh-keygen ; set timeout 5; expect \":\"; send \"\r\n\"; set timeout 3; expect \":\"; send \"\r\n\";set timeout 3; expect \":\"; send \"\r\n\"; expect eof;"

2.4.4 配置ssh免密码认证

看到过许多大牛，都是远程scp把公钥拷贝来拷贝去的。其实openssh客户端应用程序包已经提供了专门的命令来处理这个操作。

[xiaoyu@hadoop03 ~] $ ssh-copy-id -i .ssh/id_rsa.pub 192.168.88.171

The authenticity of host '192.168.88.171 (192.168.88.171)' can't be established.

RSA key fingerprint is a8:24:3f:34:86:f3:46:67:c0:a6:b0:42:86:a2:f2:c9.

Are you sure you want to continue connecting (yes/no)? yes

Warning: Permanently added '192.168.88.171' (RSA) to the list of known hosts.

Address 192.168.88.171 maps to localhost, but this does not map back to the address - POSSIBLE BREAK-IN ATTEMPT!

[email protected]'s password:

Now try logging into the machine, with "ssh '192.168.88.171'", and check in:

.ssh/authorized_keys

to make sure we haven't added extra keys that you weren't expecting.

[xiaoyu@hadoop03 ~]$

3. 在hadoop集群节点上配置

3.1 修改hosts节点

把新的集群节点的主机名和IP对应关系都放到hosts，便于本地解析
# sudo vim /etc/hosts
增加如下行：
192.168.88.171 hadoop01
192.168.88.172 hadoop02

192.168.88.173 hadoop03

可以通过ping <主机名>做简单测试

3.2 添加ssh免密码认证

方法和2.4.4 中的方法是一致的。

3.3 修改conf/slave

$ vim conf/slaves

hadoop02
hadoop03

3.4 同步配置文件到新节点上

[xiaoyu@hadoop01 hadoop-1.1.2] $ scp -r conf hadoop03:~/hadoop-1.1.2/

log4j.properties 100% 4441 4.3KB/s 00:00

capacity-scheduler.xml 100% 7457 7.3KB/s 00:00

configuration.xsl 100% 535 0.5KB/s 00:00

fair-scheduler.xml 100% 327 0.3KB/s 00:00

hdfs-site.xml 100% 319 0.3KB/s 00:00

slaves 100% 18 0.0KB/s 00:00

ssl-server.xml.example 100% 1195 1.2KB/s 00:00

hadoop-policy.xml 100% 4644 4.5KB/s 00:00

taskcontroller.cfg 100% 382 0.4KB/s 00:00

mapred-queue-acls.xml 100% 2033 2.0KB/s 00:00

ssl-client.xml.example 100% 1243 1.2KB/s 00:00

masters 100% 9 0.0KB/s 00:00

core-site.xml 100% 441 0.4KB/s 00:00

hadoop-env.sh 100% 2271 2.2KB/s 00:00

hadoop-metrics2.properties 100% 1488 1.5KB/s 00:00

mapred-site.xml 100% 261 0.3KB/s 00:00

4. 启动新节点

4.1 在新节点启动集群服务

[xiaoyu@hadoop03 hadoop-1.1.2] $ bin/hadoop-daemon.sh start datanode
starting datanode, logging to /home/xiaoyu/hadoop-1.1.2/libexec/../logs/hadoop-xiaoyu-datanode-hadoop03.out
[xiaoyu@hadoop03 hadoop-1.1.2] $ bin/hadoop-daemon.sh start tasktracker

starting tasktracker, logging to /home/xiaoyu/hadoop-1.1.2/libexec/../logs/hadoop-xiaoyu-tasktracker-hadoop03.out

5. 检查新节点的启动情况

在这里hadoop01为namenode节点

这里有三种方法，当然执行命令的方法最简便。

5.1 Namenode状态页面:http://hadoop01:50070

具体信息如下图

5.2 Jobtracker状态页面: http://hadoop01:50030

具体信息如下图：

5.3 任意节点为上执行

$ bin/hadoop dfsadmin -report

Configured Capacity: 32977600512 (30.71 GB)

Present Capacity: 20209930240 (18.82 GB)

DFS Remaining: 20003794944 (18.63 GB)

DFS Used: 206135296 (196.59 MB)

DFS Used%: 1.02%

Under replicated blocks: 1

Blocks with corrupt replicas: 0

Missing blocks: 0

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

Datanodes available: 2 (2 total, 0 dead)

Name: 192.168.88.172:50010

Decommission Status : Normal

Configured Capacity: 16488800256 (15.36 GB)

DFS Used: 205955072 (196.41 MB)

Non DFS Used: 6369054720 (5.93 GB)

DFS Remaining: 9913790464(9.23 GB)

DFS Used%: 1.25%

DFS Remaining%: 60.12%

Last contact: Fri Sep 13 03:35:51 CST 2013

Name: 192.168.88.173:50010

Decommission Status : Normal

Configured Capacity: 16488800256 (15.36 GB)

DFS Used: 180224 (176 KB)

Non DFS Used: 6398615552 (5.96 GB)

DFS Remaining: 10090004480(9.4 GB)

DFS Used%: 0%

DFS Remaining%: 61.19%

Last contact: Fri Sep 13 03:35:50 CST 2013

6. 使正在运行的计算分布到新的数据节点上

[xiaoyu@hadoop01 hadoop-1.1.2]$ ./bin/start-balancer.sh

starting balancer, logging to /home/xiaoyu/hadoop-1.1.2/libexec/../logs/hadoop-xiaoyu-balancer-hadoop01.out

[xiaoyu@hadoop01 hadoop-1.1.2]$

这个脚本很有用，大家也可以根据实际需要修改这个脚本。

7. 参考资料

Is there a way to add nodes to a running Hadoop cluster?

http://stackoverflow.com/questions/13159184/is-there-a-way-to-add-nodes-to-a-running-hadoop-cluster

1> Update the /etc/hadoop/conf/slaves list with the new node-name

2> Sync the full configuration /etc/hadoop/conf to the new datanode from the Namenode. If the file system isn't shared.

2> Restart all the hadoop services on Namenode/Tasktracker and all the services on the new Datanode.

3> Verify the new datanode from the browser http://namenode:50070

4> Run the balancer script to readjust the data between the nodes.