Hadoop - cluster

Hadoop cluster共4篇,其余Hadoop配置 - base、Hadoop配置 - hdfs、Hadoop - yarn

Prepare

server1 server2 server3 server4
NameNode y y
DataNode   y y y
ResourceManager     y y
NodeManager y y y y
Zookeeper y y y y
JournalNode y y y
ZK-FC y y    
JobHistor y
Server Name IP Host Name
server1 192.168.42.107 tdtc201
server2 192.168.42.108 tdtc202
server3 192.168.42.109 tdtc203
server4 192.168.42.110 tdtc204

1. zk

zkServer.sh start

安装详见《ZooKeeper multi-server install

2. Open firewall

Stage I: Installation

1. Unzip

/home/$USER/app/hadoop/$version

路径参考Oracle11g

mkdir -p ~/app/hadoop
tar zxvf hadoop-2.7.7.tar.gz -C ~/app/hadoop/
mv ~/app/hadoop/hadoop-2.7.7 ~/app/hadoop/2.7.7

2. set up

详见Reference[1][2][3]。

1) 再显式地重新声明一遍JAVA_HOME

hadoop-env.sh

export JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk-1.8.0.345.b01-1.el7_9.x86_64

2) set slaves

这个设置文件决定有几个DN(data node)!!!

slaves

tdtc202
tdtc203
tdtc204

3. set Hadoop Home

vi ~/.bashrc
export HADOOP_HOME=/home/tdtc/app/hadoop/2.7.7
export PATH=$PATH:$HADOOP_HOME/bin:$HADOOP_HOME/sbin
  • test
[tdtc@tdtc201 ~]$ hadoop version
Hadoop 2.7.7
Subversion Unknown -r c1aad84bd27cd79c3d1a7dd58202a8c3ee1ed3ac
Compiled by stevel on 2018-07-18T22:47Z
Compiled with protoc 2.5.0
From source with checksum 792e15d20b12c74bd6f19a1fb886490
This command was run using /home/tdtc/app/hadoop/2.7.7/share/hadoop/common/hadoop-common-2.7.7.jar

Stage II: Init

#jps

1. run journalnode

(server1,2,3)

hadoop-daemon.sh start journalnode

2. format nn

(server1 or 2)

hadoop namenode -format

1) copy file

cd ~/app/hadoop
scp -r tmp/ tdtc201:$PWD

2) format zk-fc

hdfs zkfc -formatZK

Stage III: Run

1. HDFS

server1 or 2

start-dfs.sh

stop cmd:

stop-dfs.sh

1) check

hdfs haadmin -getServiceState nn1

hdfs haadmin -getServiceState nn2

2. YARN

server3 or 4

start-yarn.sh

1) backup site

yarn-daemon.sh start resourcemanager

2) check

yarn rmadmin -getServiceState rm1

yarn rmadmin -getServiceState rm2

3. Mapreduce

server1

mr-jobhistory-daemon.sh start historyserver

web: http://192.168.42.107:50070/dfshealth.html
or
http://192.168.42.108:50070/dfshealth.html

Stage IV: Stop

$HADOOP_HOME/sbin/stop-all.sh

1. Mapreduce

mr-jobhistory-daemon.sh stop historyserver

FAQ

IP地址变更

Q: 我的IP地址变更了,怎么办?
A:

  1. 改变系统的IP

    1. /etc/hosts
    2. /etc/sysconfig/network-scripts/ifcfg-xxx
  2. 删除hadoop目录 仅限NN(name node) $rm -rf ~/app/hadoop

  3. 重新安装NN 本例子为:Server1和Server2

Reference