集群规划:
主机名 IP 安装的软件 运行的进程
Hadoop1 192.168.111.143 jdk、hadoop NameNode、DFSZKFailoverController(zkfc)、ResourceManager
Hadoop2 192.168.111.144 jdk、hadoop NameNode、DFSZKFailoverController(zkfc)、ResourceManager
Hadoop3 192.168.111.145 jdk、hadoop、zookeeper DataNode、NodeManager、JournalNode、QuorumPeerMain
Hadoop4 192.168.111.146 jdk、hadoop、zookeeper DataNode、NodeManager、JournalNode、QuorumPeerMain
Hadoop5 192.168.111.147 jdk、hadoop、zookeeper DataNode、NodeManager、JournalNode、QuorumPeerMain
1. zookeeper集群搭建
1.1 解压
tar -zxvf zookeeper-3.4.9.tar.gz -C /home/hbase
1.2 修改配置
cd /home/hbase/zookeeper-3.4.9/conf/
cp zoo_sample.cfg zoo.cfg
vim zoo.cfg
修改:
dataDir=/home/hbase/zookeeper-3.4.9/tmp
在zoo.cfg最后添加:
server.1=hadoop3:2888:3888
server.2=hadoop4:2888:3888
server.3=hadoop5:2888:3888
然后创建一个tmp文件夹
mkdir /home/hbase/zookeeper-3.4.9/tmp
再创建一个空文件
touch /home/hbase/zookeeper-3.4.9/tmp/myid
最后向该文件写入ID
echo 1 >> /home/hbase/zookeeper-3.4.9/tmp/myid
1.3 将配置好的zookeeper拷贝到其他节点
scp -r /home/hbase/zookeeper-3.4.9/ hadoop4: /home/hbase/
scp -r /home/hbase/zookeeper-3.4.9/ hadoop5: /home/hbase/
注意:修改hadoop4、hadoop5对应/home/hbase /zookeeper-3.4.9/tmp/myid内容
hadoop4:
echo 2 >> /home/hbase/zookeeper-3.4.9/tmp/myid
hadoop5:
echo 3 >> /home/hbase/zookeeper-3.4.9/tmp/myid
2. 安装配置hadoop集群(在hadoop1上操作)
2.1 解压
tar -zxvf hadoop-2.7.2.tar.gz -C /home/hbase/
2.2 配置HDFS
#将hadoop添加到环境变量中
vim /etc/profile
export JAVA_HOME=/home/habse/jdk/jdk1.7.0_79
export HADOOP_HOME=/home/habse/hadoop-2.7.2
export PATH=$PATH:$JAVA_HOME/bin:$HADOOP_HOME/bin
#hadoop2.0的配置文件全部在$HADOOP_HOME/etc/hadoop下
cd /home/habse/hadoop-2.7.2/etc/hadoop
2.2.1 修改hadoop-env.sh
export JAVA_HOME=/home/hbase/jdk/jdk1.7.0_79
2.2.2 修改core-site.xml
fs.defaultFS
hdfs://ns1
hadoop.tmp.dir
/home/habse/hadoop-2.7.2/tmp
ha.zookeeper.quorum
hadoop3:2181,hadoop4:2181,hadoop5:2181
2.2.3 修改hdfs-site.xml
dfs.nameservices
ns1
dfs.ha.namenodes.ns1
nn1,nn2
dfs.namenode.rpc-address.ns1.nn1
hadoop1:9000
dfs.namenode.http-address.ns1.nn1
hadoop1:50070
dfs.namenode.rpc-address.ns1.nn2
hadoop2:9000
dfs.namenode.http-address.ns1.nn2
hadoop2:50070
dfs.namenode.shared.edits.dir
qjournal://hadoop3:8485;hadoop4:8485;hadoop5:8485/ns1
dfs.journalnode.edits.dir
/home/hbase/hadoop-2.7.2/journal
dfs.ha.automatic-failover.enabled
true
dfs.client.failover.proxy.provider.ns1
org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider
dfs.ha.fencing.methods
sshfence
shell(/bin/true)
dfs.ha.fencing.ssh.private-key-files
/root/.ssh/id_rsa
dfs.ha.fencing.ssh.connect-timeout
30000
2.2.4 修改mapred-site.xml
mapreduce.framework.name
yarn
2.2.5 修改yarn-site.xml
yarn.resourcemanager.ha.enabled
true
yarn.resourcemanager.cluster-id
yrc
yarn.resourcemanager.ha.rm-ids
rm1,rm2
yarn.resourcemanager.hostname.rm1
hadoop1
yarn.resourcemanager.hostname.rm2
hadoop2
yarn.resourcemanager.zk-address
hadoop3:2181,hadoop4:2181,hadoop5:2181
yarn.nodemanager.aux-services
mapreduce_shuffle
2.2.6 修改slaves
slaves是指定子节点的位置, hadoop1上的slaves文件指定的是datanode和nodemanager的位置
hadoop3
hadoop4
hadoop5
2.2.7 配置免密码登陆
#首先要配置hadoop1到hadoop2、hadoop3、hadoop4、hadoop5的免密码登陆
#在hadoop1上生产一对钥匙
ssh-keygen -t rsa
#将公钥拷贝到其他节点,包括自己
ssh-coyp-id hadoop1
ssh-coyp-id hadoop2
ssh-coyp-id hadoop3
ssh-coyp-id hadoop4
ssh-coyp-id hadoop5
#注意:两个namenode之间要配置ssh免密码登陆,别忘了配置hadoop2到hadoop1的免登陆
在hadoop2上生产一对钥匙
ssh-keygen -t rsa
ssh-coyp-id -i hadoop1
2.3 将配置好的hadoop拷贝到其他节点
scp -r /home/habse/hadoop-2.7.2/ root@hadoop2:/home/habse/
scp -r /home/habse/hadoop-2.7.2/ root@hadoop3:/home/habse /
scp -r /home/habse/hadoop-2.7.2/ root@hadoop4:/home/habse /
scp -r /home/habse/hadoop-2.7.2/ root@hadoop5:/home/habse /
3. 第一次启动
3.1 启动zookeeper集群(分别在hadoop3、hadoop4、hadoop5上启动zk)
cd /home/hbase/zookeeper-3.4.9/bin/
./zkServer.sh start
#查看状态:一个leader,两个follower
./zkServer.sh status
3.2 启动journalnode(分别在在hadoop3、hadoop4、hadoop5上执行)
cd /home/habse/hadoop-2.7.2
sbin/hadoop-daemon.sh start journalnode
#运行jps命令检验,hadoop3、hadoop4、hadoop5上多了JournalNode进程
3.3 格式化HDFS
#在hadoop1上执行命令:
hdfs namenode -format
hdfs namenode -bootstrapStandby
3.4 格式化ZK(在hadoop1上执行即可)
hdfs zkfc -formatZK
3.5 启动HDFS(在hadoop1上执行)
sbin/start-dfs.sh
注意:
如果启动datanode时遇到找不到datanode所在的主机,首先检查slaves文件配置是否正确,如果没问题的话删除重新创建
3.6 启动YARN(hadoop1上执行)
sbin/start-yarn.sh
查看每台机器的进程:
到此,hadoop-2.7.2配置完毕,可以统计浏览器访问:
http://192.168.111.143:50070
NameNode 'hadoop1:9000' (active)
http://192.168.111.144:50070
NameNode 'hadoop2:9000' (standby)
Datanode:
所以hadoop集群安装完成后首先启动zookeeper和journalnode,然后格式化HDFS和ZKFC,然后启动namenode,resourcemanager,datanode
4. startup与shutdown
4.1 Hadoop Startup
1. ./zkServer.sh start(hadoop3、hadoop4、hadoop5)
2. ./hadoop-daemon.sh start journalnode(hadoop3、hadoop4、hadoop5)
3. hdfs zkfc -formatZK(hadoop1)
4. hdfs namenode -bootstrapStandby(hadoop2)
5. hdfs zkfc -formatZK(hadoop1)
6. ./start-dfs.sh (hadoop1)
7. ./start-yarn.sh(hadoop1)
8. 如果哪个进程没有启动,那么单独在那台机器上执行启动命令
9. ./yarn –daemon start proxyserver
10. ./mapred –daemon start historyserver
说明:
格式化工作仅在第一次启动hadoop之前完成(步骤2,3,4,5),以后不用,如果以后启动过程中有问题可以重新格式化
单独启动resourcemanager:./yarn-daemon.sh start resourcemanager
单独启动namnode:./hadoop-daemon.sh start namenode
单独启动zkfc:./yarn-daemon.sh start zkfc
4.2 Hadoop shutdown
1. ./stop-dfs.sh
2. ./stop-yarn.sh
3. ./yarn –daemon stop proxyserver
4. ./mapred –daemon stop historyserver
5. 主备切换测试
杀掉当前状态为active的hadoop1 的namenode进程,可以看到hadoop2由standby变为active,再启动hadoop1的namenode则发现hadoop1的状态为standby