Apache Hadoop 集群安装文档
简介：
Apache Hadoop 集群安装文档
软件：jdk-8u111-linux-x64.rpm、hadoop-2.8.0.tar.gz
http://www.apache.org/dyn/closer.cgi/hadoop/common/hadoop-2.8.0/hadoop-2.8.0.tar.gz
系统：CentOS 6.8 x64 主机列表及配置信息： master.hadoop datanode[01:03].hadoop CPU： 8 4 MEM: 16G 8G DISK： 100G*2 100G*2
一、系统初始化
# master.hadoop
shell > vim /etc/hosts 192.168.1.25 master.hadoop 192.168.1.27 datanode01.hadoop 192.168.1.28 datanode02.hadoop 192.168.1.29 datanode03.hadoop shell > yum -y install epel-release shell > yum -y install ansible shell > ssh-keygen # 生成密钥 shell > ssh-copy-id -i ~/.ssh/id_rsa.pub "-p 22 root@datanode01.hadoop" shell > ssh-copy-id -i ~/.ssh/id_rsa.pub "-p 22 root@datanode02.hadoop" shell > ssh-copy-id -i ~/.ssh/id_rsa.pub "-p 22 root@datanode03.hadoop" shell > vim /etc/ansible/hosts # datanode.hadoop [datanode] datanode[01:03].hadoop shell > ansible datanode -m shell -a 'useradd hadoop && echo hadoop | passwd --stdin hadoop' shell > ansible datanode -m shell -a "echo '* - nofile 65536' >> /etc/security/limits.conf" shell > ansible datanode -m copy -a 'src=/etc/hosts dest=/etc/hosts' # 同步 hosts shell > ansible datanode -m shell -a '/etc/init.d/iptables stop && chkconfig --del iptables' # 关闭防火墙 shell > ansible datanode -m shell -a 'sed -i '/SELINUX/s/enforcing/disabled/' /etc/selinux/config' # 关闭 SELinux shell > ansible datanode -m shell -a 'echo 'vm.swappiness = 0' >> /etc/sysctl.conf' # 修改内核参数 shell > ansible datanode -m shell -a 'echo 'echo never > /sys/kernel/mm/redhat_transparent_hugepage/defrag' # 关闭透明大页 shell > ansible datanode -m shell -a 'echo 'echo never > /sys/kernel/mm/redhat_transparent_hugepage/defrag' >> /etc/rc.local' shell > ansible datanode -m shell -a 'reboot'
# 上面的 ansible 操作，master.hadoop 也要执行
二、时间同步
# master.hadoop
shell > /bin/cp -f /usr/share/zoneinfo/Asia/Shanghai /etc/localtime shell > yum -y install ntp shell > /etc/init.d/ntpd stop && chkconfig --del ntpd shell > ntpdate us.pool.ntp.org | hwclock -w shell > vim /etc/ntp.conf # 允许时间同步客户端 restrict 192.168.1.0 mask 255.255.255.0 nomodify # Server 向谁同步时间 server us.pool.ntp.org prefer # Server 无法向时间服务器同步时，使用本地时钟 server 127.127.1.0 fudge 127.127.1.0 stratum 10 shell > /etc/init.d/ntpd start shell > echo -e '\n/usr/sbin/ntpdate us.pool.ntp.org | hwclock -w > /dev/null' >> /etc/rc.local shell > echo -e '\n/etc/init.d/ntpd start > /dev/null' >> /etc/rc.local shell > ansible datanode -m shell -a 'yum -y install ntpdate' shell > ansible datanode -m shell -a '/bin/cp -f /usr/share/zoneinfo/Asia/Shanghai /etc/localtime' shell > ansible datanode -m shell -a 'ntpdate master.hadoop | hwclock -w' shell > ansible datanode -m cron -a "name='ntpdate master.hadoop' minute=0 hour=0 job='/usr/sbin/ntpdate master.hadoop | hwclock -w > /dev/null'"
三、集群部署
# master.hadoop
1、安装 jdk、下载、解压 apache hadoop、设置主机间 hadoop 用户无密码登录
shell > rpm -ivh /usr/local/src/jdk-8u111-linux-x64.rpm shell > echo 'export JAVA_HOME=/usr/java/default' >> /etc/profile && source /etc/profile shell > tar zxf /usr/local/src/hadoop-2.8.0.tar.gz -C /usr/local/ shell > chown -R hadoop.hadoop /usr/local/hadoop-2.8.0 shell > echo -e '\nexport PATH=$PATH:/usr/local/hadoop-2.8.0/bin' >> /etc/profile && source /etc/profile shell > su - hadoop hadoop shell > ssh-keygen hadoop shell > cat .ssh/id_rsa.pub > .ssh/authorized_keys && chmod 600 .ssh/authorized_keys hadoop shell > ssh-copy-id -i ~/.ssh/id_rsa.pub "-p 22 hadoop@datanode01.hadoop" hadoop shell > ssh-copy-id -i ~/.ssh/id_rsa.pub "-p 22 hadoop@datanode02.hadoop" hadoop shell > ssh-copy-id -i ~/.ssh/id_rsa.pub "-p 22 hadoop@datanode03.hadoop"
2、配置 apache hadoop
# 指定 Slave、即 DataNode、NodeManager 角色
hadoop shell > vim /usr/local/hadoop-2.8.0/etc/hadoop/slaves datanode01.hadoop datanode02.hadoop datanode03.hadoop
# 修改 hadoop-env.sh
hadoop shell > vim /usr/local/hadoop-2.8.0/etc/hadoop/hadoop-env.sh export JAVA_HOME=/usr/java/default
# 修改 core-site.xml
<configuration>
<property>
<name>fs.defaultFS</name>
<value>hdfs://master.hadoop:8020</value>
</property>
<property>
<name>hadoop.tmp.dir</name>
<value>file:///data/hadoop/tmp</value>
</property>
<property>
<name>fs.trash.interval</name>
<value>1440</value>
</property>
<property>
<name>io.file.buffer.size</name>
<value>131072</value>
</property>
</configuration>
# hadoop 核心配置文件
# 默认加载项 HADOOP_HOME/share/doc/hadoop/hadoop-project-dist/hadoop-common/core-default.xml
# fs.defaultFS NameNode IP:PORT，老版本为 fs.default.name
# hadoop.tmp.dir hadoop 临时目录，很多目录不明确配置时，都基于该目录 ( 默认 /tmp，系统重启时会被删除 )，很重要！
# fs.trash.interval 开启垃圾回收，1440 分钟，默认 0 关闭 ( 用户文件系统级删除的数据会被移到回收站，24小时后被删除 )
# io.file.buffer.size 读写流文件缓存大小，减少IO次数，默认 4096 字节
# 修改 hdfs-site.xml
hadoop shell > vim /usr/local/hadoop-2.8.0/etc/hadoop/hdfs-site.xml <configuration> <property> <name>dfs.blocksize</name> <value>134217728</value> </property> <property> <name>dfs.replication</name> <value>3</value> </property> <property> <name>dfs.namenode.name.dir</name> <value>file:///data/dfs/nn</value> </property> <property> <name>dfs.namenode.checkpoint.dir</name> <value>file:///data/dfs/sn</value> </property> <property> <name>dfs.datanode.data.dir</name> <value>file:///data/dfs/dn</value> </property> <property> <name>dfs.namenode.handler.count</name> <value>20</value> </property> </configuration>
# HDFS 配置文件
# 默认加载项 HADOOP_HOME/share/doc/hadoop/hadoop-project-dist/hadoop-hdfs/hdfs-default.xml
# dfs.hosts / dfs.hosts.exclude 允许或排除某些 DataNode 连接 NameNode
# dfs.blocksize 块大小，默认 134217728 ( 128M )
# dfs.replication 默认副本数，数据冗余
# dfs.namenode.name.dir NameNode 元数据存放位置，可以配置多个目录，以 , 分割，用作数据冗余！
# dfs.namenode.checkpoint.dir SecondaryNameNode 数据存储目录，该角色负责将 NameNode 的 edit log 合并到 fsimage
# dfs.datanode.data.dir DataNode 数据存放位置，可以配置多个目录，以 , 分割，数据轮询写入，增加写入速度 ( 多个目录应该对应多个设备 DISK )
# dfs.namenode.handler.count NameNode 线程数，用于跟 DataNode 通信，默认 10，增大该参数可以优化性能，但是资源也相应提升
# 修改 yarn-site.xml
hadoop shell > vim /usr/local/hadoop-2.8.0/etc/hadoop/yarn-site.xml <configuration> <property> <name>yarn.resourcemanager.hostname</name> <value>master.hadoop</value> </property> <property> <name>yarn.resourcemanager.scheduler.class</name> <value>org.apache.hadoop.yarn.server.resourcemanager.scheduler.capacity.CapacityScheduler</value> </property> <property> <name>yarn.nodemanager.log-dirs</name> <value>${yarn.log.dir}/userlogs</value> </property> <property> <name>yarn.nodemanager.remote-app-log-dir</name> <value>/tmp/logs</value> </property> <property> <name>yarn.nodemanager.aux-services</name> <value>mapreduce_shuffle</value> </property> </configuration>
# YARN 配置文件
# 默认加载项 HADOOP_HOME/share/doc/hadoop/hadoop-yarn/hadoop-yarn-common/yarn-default.xml
# yarn.resourcemanager.hostname ReSourceManager 主机，其余关于端口的监听都基于该配置项
# yarn.resourcemanager.scheduler.class 资源调度算法，CapacityScheduler 计算能力调度、FairScheduler 公平调度、Fifo Scheduler 先进先出调度
# yarn.nodemanager.log-dirs nodemanager 日志目录
# yarn.nodemanager.remote-app-log-dir nodemanager 中间结果保持目录
# 修改 mapred-site.xml
hadoop shell > cat /usr/local/hadoop-2.8.0/etc/hadoop/mapred-site.xml.template > /usr/local/hadoop-2.8.0/etc/hadoop/mapred-site.xml hadoop shell > vim /usr/local/hadoop-2.8.0/etc/hadoop/mapred-site.xml <configuration> <property> <name>mapreduce.framework.name</name> <value>yarn</value> </property> <property> <name>mapreduce.jobhistory.address</name> <value>master.hadoop:10020</value> </property> <property> <name>mapreduce.jobhistory.webapp.address</name> <value>master.hadoop:19888</value> </property> <property> <name>yarn.app.mapreduce.am.staging-dir</name> <value>/tmp/hadoop-yarn/staging</value> </property> </configuration>
# MAPREDUCE 配置文件
# 默认加载项 HADOOP_HOME/share/doc/hadoop/hadoop-mapreduce-client/hadoop-mapreduce-client-core/mapred-default.xml
# mapreduce.framework.name 使用 yarn 来管理资源
# yarn.app.mapreduce.am.staging-dir 提交作业时的临时目录，提交作业历史目录 mapreduce.jobhistory.done-dir、mapreduce.jobhistory.intermediate-done-dir 都基于该目录
hadoop shell > exit
3、部署 Slave
shell > ansible datanode -m copy -a 'src=/usr/local/src/jdk-8u111-linux-x64.rpm dest=/usr/local/src/' shell > yum -y install rsync shell > ansible datanode -m shell -a 'yum -y install rsync' shell > ansible datanode -m synchronize -a 'src=/usr/local/hadoop-2.8.0 dest=/usr/local/' # 我还傻傻的用 copy 模块，结果慢的要死，synchroize 为 rsync 模块，好快！ shell > ansible datanode -m shell -a 'rpm -ivh /usr/local/src/jdk-8u111-linux-x64.rpm' shell > ansible datanode -m shell -a "echo -e '\nexport JAVA_HOME=/usr/java/default' >> /etc/profile && source /etc/profile"
四、启动集群
# master.hadoop
shell > chmod -R a+w /data shell > ansible datanode -m shell -a 'chmod -R a+w /data' # 需要给 /data 目录写入权限，否则无法初始化文件系统 hdfs namenode -format shell > su - hadoop hadoop shell > hdfs namenode -format # 初次启动需要格式化文件系统 hadoop shell > sh /usr/local/hadoop-2.8.0/sbin/start-all.sh # 启动所有服务 / stop-all.sh 关闭服务 hadoop shell > jps 4386 ResourceManager 4659 Jps 3990 NameNode 4204 SecondaryNameNode
# 这是 master.hadoop 启动的角色
# http://192.168.1.25:50070 # NameNode
# http://192.168.1.25:8088 # ReSourceManagerv
# http://192.168.1.25:10020 # MapReduce JobHistory Server :19888 webui
# datanode.hadoop
hadoop shell > jps 2508 Jps 2238 DataNode 2351 NodeManager
# 这是 datanode.hadoop 启动的角色
hadoop shell > hdfs dfs -ls ls: `.': No such file or directory hadoop shell > hdfs dfs -mkdir /user hadoop shell > hdfs dfs -mkdir /user/hadoop hadoop shell > hdfs dfs -ls
# 为 hadoop 用户创建家目录
五、运行示例
# master.hadoop
hadoop shell > hdfs dfs -put shakespeare.txt # 上传本地文件到 hdfs hadoop shell > hdfs dfs -ls Found 1 items -rw-r--r-- 3 hadoop supergroup 5447165 2017-05-17 16:49 shakespeare.txt hadoop shell > hadoop jar /usr/local/hadoop-2.8.0/share/hadoop/mapreduce/hadoop-mapreduce-examples-2.8.0.jar grep shakespeare.txt outfile what # 执行官方示例，词频统计 ( 统计 what 出现次数 ) hadoop shell > hdfs dfs -ls drwxr-xr-x - hadoop supergroup 0 2017-04-11 19:38 outfile -rw-r--r-- 3 hadoop supergroup 5447165 2017-04-11 19:35 shakespeare.txt hadoop shell > hdfs dfs -cat outfile/* 2309 what
报错管理：
1、bin/hdfs namenode -format # 初始化文件系统报错
17/04/01 19:04:29 ERROR namenode.NameNode: Failed to start namenode. java.io.IOException: Cannot create directory /data/dfs/namenode/current at org.apache.hadoop.hdfs.server.common.Storage$StorageDirectory.clearDirectory(Storage.java:352) at org.apache.hadoop.hdfs.server.namenode.NNStorage.format(NNStorage.java:573) at org.apache.hadoop.hdfs.server.namenode.NNStorage.format(NNStorage.java:594) at org.apache.hadoop.hdfs.server.namenode.FSImage.format(FSImage.java:156) at org.apache.hadoop.hdfs.server.namenode.NameNode.format(NameNode.java:1102) at org.apache.hadoop.hdfs.server.namenode.NameNode.createNameNode(NameNode.java:1544) at org.apache.hadoop.hdfs.server.namenode.NameNode.main(NameNode.java:1671)
# 解决方法
shell > chmod -R a+w /data shell > ansible datanode -m shell -a 'chmod -R a+w /data'
2、WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable # 迷之警告
