CDH4.0安装及配置(三)

一.CDH4安装
在官网https://ccp.cloudera.com/display/CDH4DOC/CDH4+Installation下载针对本机操作系统的CDH4引导包,这里我的操作系统是redhat5.4
Step 1a: Optionally Add a Repository Key
    rpm --import http://archive.cloudera.com/cdh4/redhat/5/x86_64/cdh/RPM-GPG-KEY-cloudera
Step 2: Install CDH4 with MRv1
    yum -y install hadoop-0.20-mapreduce-jobtracker
Step 3: Install CDH4 with YARN
    yum -y install hadoop-yarn-resourcemanager
 yum -y install hadoop-hdfs-namenode
 yum -y install hadoop-hdfs-secondarynamenode
 yum -y install hadoop-yarn-nodemanager hadoop-hdfs-datanode hadoop-mapreduce
 yum -y install hadoop-mapreduce-historyserver hadoop-yarn-proxyserver
 yum -y install hadoop-client

另:请实现装好jdk,Postgresql

二.CDH4配置
1.配置网络主机
(1).配置网络主机
    为确保主机之间可以互信,/etc/hosts内容,/etc/sysconfig/network
 注意跟ip地址保持一直
(2).复制hadoop配置
    cp -r /etc/hadoop/conf.empty /etc/hadoop/conf.my_cluster
(3).自定义配置文件
    /etc/hadoop/conf/core-site.xml
     fs.default.name(老版的,已弃用,但仍然兼容可用) or fs.defaultFS指定namenode的文件系统
  例:
     <property>
               <name>fs.defaultFS</name>
               <value>hdfs://namenode-host.company.com/</value>
           </property>
 /etc/hadoop/conf/hdfs-site.xml
     dfs.permissions.superusergroup指定的UNIX组包含用户，将被视为由HDFS的超级用户
     例:
     <property>
               <name>dfs.permissions.superusergroup</name>
               <value>hadoop</value>
           </property>
(4).配置本地存储目录
    ①./etc/hadoop/conf/hdfs-site.xml
   namenode:
     dfs.name.dir(老版的,已弃用,但仍然兼容可用)或dfs.namenode.name.dir此属性指定的目录，其中NameNode的存储元数据和编辑日志。Cloudera的建议指定至少两个目录，其中之一是位于一个NFS挂载点
     例:
        <property>
                  <name>dfs.namenode.name.dir</name>
                  <value>/data/1/dfs/nn,/nfsmount/dfs/nn</value>
              </property>
   datanode:
     dfs.data.dir(老版的,已弃用,但仍然兼容可用)或dfs.datanode.data.dir此属性指定目录下的DataNode块存储。Cloudera的建议配置一个独立的磁盘,挂载到上面
     例:       
              <property>
                  <name>dfs.datanode.data.dir</name>
                  <value>/data/1/dfs/dn,/data/2/dfs/dn,/data/3/dfs/dn</value>
              </property>
 ②.创建上面所用到的目录
 mkdir -p /data/1/dfs/nn /nfsmount/dfs/nn
 mkdir -p /data/1/dfs/dn /data/2/dfs/dn /data/3/dfs/dn /data/4/dfs/dn
 chown -R hdfs:hdfs /data/1/dfs/nn /nfsmount/dfs/nn /data/1/dfs/dn /data/2/dfs/dn /data/3/dfs/dn /data/4/dfs/dn
 ③.最后正确的权限是
 dfs.name.dir or or dfs.namenode.name.dir | hdfs:hdfs | drwx------
 -----------------------------------------------------------------
 dfs.data.dir or dfs.datanode.data.dir    | hdfs:hdfs | drwx------
 ④.注:Hadoop守护进程会自动为你设置正确的权限dfs.data.dir或dfs.datanode.data.dir。但在dfs.name.dir或dfs.namenode.name.dir的情况下，权限是当前不正确的设置为默认的文件系统，通常是drwxr - XR的-X（755）。使用命令将dfs.name.dir或dfs.namenode.name.dir目录权限设置为drwx ------
      chmod 700 /data/1/dfs/nn /nfsmount/dfs/nn
   chmod go-rx /data/1/dfs/nn /nfsmount/dfs/nn
(5).格式化namenode
service hadoop-hdfs-namenode init
(6).配置远程NameNode的存储目录
mount -t nfs -o tcp,soft,intr,timeo=10,retrans=10, <server>:<export> <mount_point>
如果针对的是一个HA高可用集群的话,则
mount -t nfs -o tcp,soft,intr,timeo=50,retrans=12, <server>:<export> <mount_point>     
2.在集群上部署mapreduce的MRv1集群
    (1).Step 1: Configuring Properties for MRv1 Clusters
 /etc/hadoop/conf/mapred-site.xml
 mapred.job.tracker指定JobTracker的RPC服务器的主机名和（可选）端口<HOST>：<PORT>,指定的必须是主机而不是ip地址
 例:
    <property>
           <name>mapred.job.tracker</name>
           <value>jobtracker-host.company.com:8021</value>
       </property>
 (2).Step 2: Configure Local Storage Directories for Use by MRv1 Daemons
 /etc/hadoop/conf/mapred-site.xml
 mapred.local.dir指定用于存放临时数据和中间文件的目录
 例:
    <property>
           <name>mapred.local.dir</name>
           <value>/data/1/mapred/local,/data/2/mapred/local,/data/3/mapred/local</value>
       </property>
 创建这几个目录
 mkdir -p /data/1/mapred/local /data/2/mapred/local /data/3/mapred/local /data/4/mapred/local
 配置属主属组
 chown -R mapred:hadoop /data/1/mapred/local /data/2/mapred/local /data/3/mapred/local /data/4/mapred/local
 (3).Step 3: Configure a Health Check Script for DataNode Processes
 健康检查,这里提供一个官方的脚本
 #!/bin/bash
    if ! jps | grep -q DataNode ; then
      echo ERROR: datanode not up
    fi
 (4).Step 4: Deploy your Custom Configuration to your Entire Cluster
 设置每个节点的配置文件
 alternatives --set hadoop-conf /etc/hadoop/conf.my_cluster
 (5).Step 5: Start HDFS
 for service in /etc/init.d/hadoop-hdfs-*
    > do
    > sudo $service start
    > done
 (6).Step 6: Create the HDFS /tmp Directory
 sudo -u hdfs hadoop fs -mkdir /tmp
 sudo -u hdfs hadoop fs -chmod -R 1777 /tmp
 注意:这是使用本地文件系统HDFS，这是根的hadoop.tmp.dir
 (7).Step 7: Create MapReduce /var directories
 sudo -u hdfs hadoop fs -mkdir /var
    sudo -u hdfs hadoop fs -mkdir /var/lib
    sudo -u hdfs hadoop fs -mkdir /var/lib/hadoop-hdfs
    sudo -u hdfs hadoop fs -mkdir /var/lib/hadoop-hdfs/cache
    sudo -u hdfs hadoop fs -mkdir /var/lib/hadoop-hdfs/cache/mapred
    sudo -u hdfs hadoop fs -mkdir /var/lib/hadoop-hdfs/cache/mapred/mapred
    sudo -u hdfs hadoop fs -mkdir /var/lib/hadoop-hdfs/cache/mapred/mapred/staging
    sudo -u hdfs hadoop fs -chmod 1777 /var/lib/hadoop-hdfs/cache/mapred/mapred/staging
    sudo -u hdfs hadoop fs -chown -R mapred /var/lib/hadoop-hdfs/cache/mapred
 (8).Step 8: Verify the HDFS File Structure
 检查hdfs文件结构
 sudo -u hdfs hadoop fs -ls -R /
 (9).Step 9: Create and Configure the mapred.system.dir Directory in HDFS
 ①.sudo -u hdfs hadoop fs -mkdir /mapred/system
 sudo -u hdfs hadoop fs -chown mapred:hadoop /mapred/system
 ②.正确的权限
 mapred.system.dir | mapred:hadoop | drwx------ 1
 /                 | hdfs:hadoop   | drwxr-xr-x
 (10).Step 10: Start MapReduce
 在TaskTracker系统
 sudo service hadoop-0.20-mapreduce-tasktracker start
 在JobTracker系统
 sudo service hadoop-0.20-mapreduce-jobtracker start
 (11).Step 11: Create a Home Directory for each MapReduce User
 创建每个MapReduce用户的主目录
 sudo -u hdfs hadoop fs -mkdir  /user/<user>
 sudo -u hdfs hadoop fs -chown <user> /user/<user>