Spark修炼之道（进阶篇）——Spark入门到精通：第一节 Spark 1.5.0集群搭建
作者：周志湖
网名：摇摆少年梦
微信号：zhouzhihubeyond
本节主要内容
- 操作系统环境准备
- Hadoop 2.4.1集群搭建
- Spark 1.5.0 集群部署
注：在利用CentOS 6.5操作系统安装spark 1.5集群过程中，本人发现Hadoop 2.4.1集群可以顺利搭建，但在Spark 1.5.0集群启动时出现了问题（可能原因是64位操作系统原因，源码需要重新编译，但本人没经过测试），经本人测试在ubuntu 10.04 操作系统上可以顺利成功搭建。大家可以利用CentOS 6.5进行尝试，如果有问题，再利用ubuntu 10.04搭建，所有步骤基本一致
1. 操作系统环境准备
（1）安装VMWare
下载地址：http://pan.baidu.com/s/1bniBipD
密码：pbdw
安装过程略
（2）下载操作系统并安装
Ubuntu 10.04操作系统下载地址：
链接：http://pan.baidu.com/s/1kTy9Umj 密码：2w5bCentOS 6.5下载地址：
下载地址：http://pan.baidu.com/s/1mgkuKdi
密码：xtm5本实验要求装三台：CentOS 6.5，可以分别安装，也可以安装完一台后克隆两台，具体过程略。初学者，建议三台分别安装。安装后如下图所示： 
（3）CentOS 6.5网络配置
安装好的虚拟机一般默认使用的是NAT（关于NAT、桥接等虚拟机网络连接方式参见本人博客：http://blog.csdn.net/lovehuangjiaju/article/details/48183485），由于三台机器之间需要互通之外，还需要与本机连通，因此采用将网络连接方式设置为Bridged（三台机器相同的设置），如下图所法： 
修改主机名
（1）修改centos_salve01虚拟机主机名：
vim /etc/sysconfig/network/etc/sysconfig/network修改后的内容如下： 
（2）vim /etc/sysconfig/network命令修改centos_slave02虚拟机主机名
/etc/sysconfig/network修改后的内容如下： 
（3）vim /etc/sysconfig/network命令修改centos_slave03虚拟机主机名
/etc/sysconfig/network修改后的内容如下： 
修改主机IP地址
在大家在配置时，修改/etc/sysconfig/network-scripts/ifcfg-eth0文件对应的BOOTPROT=static、IPADDR、NETMASK、GATEWAY及DNS1信息即可
（1）修改centos_salve01虚拟机主机IP地址：
vim /etc/sysconfig/network-scripts/ifcfg-eth0- 1
修改后内容如下：
DEVICE="eth0"
BOOTPROTO="static"
HWADDR="00:0c:29:3f:69:4d"
IPV6INIT="yes"
NM_CONTROLLED="yes"
ONBOOT="yes"
TYPE="Ethernet"
UUID="5315276c-db0d-4061-9c76-9ea86ba9758e"
IPADDR="192.168.1.111"
NETMASK="255.255.255.0"
GATEWAY="192.168.1.1"
DNS1="8.8.8.8"
（2）修改centos_salve02虚拟机主机IP地址：
vim /etc/sysconfig/network-scripts/ifcfg-eth0- 1
修改后内容如下：
DEVICE="eth0"
BOOTPROTO="static"
HWADDR="00:0c:29:64:f9:80"
IPV6INIT="yes"
NM_CONTROLLED="yes"
ONBOOT="yes"
TYPE="Ethernet"
UUID="5315276c-db0d-4061-9c76-9ea86ba9758e"
IPADDR="192.168.1.112"
NETMASK="255.255.255.0"
GATEWAY="192.168.1.1"
DNS1="8.8.8.8"
（3）修改centos_salve03虚拟机主机IP地址：
vim /etc/sysconfig/network-scripts/ifcfg-eth0- 1
修改后内容如下：
DEVICE="eth0"
BOOTPROTO="static"
HWADDR="00:0c:29:1e:80:b1"
IPV6INIT="yes"
NM_CONTROLLED="yes"
ONBOOT="yes"
TYPE="Ethernet"
UUID="5315276c-db0d-4061-9c76-9ea86ba9758e"
IPADDR="192.168.1.113"
NETMASK="255.255.255.0"
GATEWAY="192.168.1.1"
DNS1="8.8.8.8"
/etc/sysconfig/network-scripts/ifcfg-eth0文件内容解析：
DEVICE=eth0 //指出设备名称
BOOTPROT=static //启动类型 dhcp|static，使用桥接模式，必须是static
HWADDR=00:06:5B:FE:DF:7C //硬件Mac地址
IPADDR=192.168.0.2 //IP地址
NETMASK=255.255.255.0 //子网掩码
NETWORK=192.168.0.0 //网络地址
GATEWAY=192.168.0.1 //网关地址
ONBOOT=yes //是否启动应用
TYPE=Ethernet //网络类型设置完成后，使用
service network restart- 1
命令重新启动网络，配置即可生效。
设置主机名与IP地址映射
（1）修改centos_salve01主机名与IP地址映射
vim /etc/hosts- 1
设置内容如下：
127.0.0.1 slave01.example.com localhost localhost.localdomain localhost4 localhost4.localdomain4
::1 slave01.example.com
192.168.1.111 slave01.example.com
192.168.1.112 slave02.example.com
192.168.1.113 slave03.example.com
具体如下图：
（2）修改centos_salve02主机名与IP地址映射
vim /etc/hosts设置内容如下：
127.0.0.1 slave02.example.com localhost localhost.localdomain localhost4 localhost4.localdomain4
::1 slave02.example.com
192.168.1.111 slave01.example.com
192.168.1.112 slave02.example.com
192.168.1.113 slave03.example.com
具体如下图： 
（3）修改centos_salve03主机名与IP地址映射
vim /etc/hosts设置内容如下：
127.0.0.1 slave03.example.com localhost localhost.localdomain localhost4 localhost4.localdomain4
::1 slave03.example.com
192.168.1.111 slave01.example.com
192.168.1.112 slave02.example.com
192.168.1.113 slave03.example.com
修改主机DNS
采用下列命令设置各主机DNS（三台机器进行相同的设置)
vim /etc/resolv.conf 设置后的内容：
# Generated by NetworkManager
search example.com
nameserver 8.8.8.8
8.8.8.8为Google提供的DNS服务器
网络连通测试
前面所有的配置完成后，重启centos_salve01、centos_salve02、centos_salve03使主机名设置生效，然后分别在三台机器上作如下测试命令：
下面只给出在centos_salve01虚拟机上的测试
[root@slave01 ~]# ping slave02.example.com
PING slave02.example.com (192.168.1.112) 56(84) bytes of data.
64 bytes from slave02.example.com (192.168.1.112): icmp_seq=1 ttl=64 time=0.417 ms
64 bytes from slave02.example.com (192.168.1.112): icmp_seq=2 ttl=64 time=0.355 ms
64 bytes from slave02.example.com (192.168.1.112): icmp_seq=3 ttl=64 time=0.363 ms
^C
--- slave02.example.com ping statistics ---
3 packets transmitted, 3 received, 0% packet loss, time 2719ms
rtt min/avg/max/mdev = 0.355/0.378/0.417/0.031 ms
[root@slave01 ~]# ping slave03.example.com
PING slave03.example.com (192.168.1.113) 56(84) bytes of data.
64 bytes from slave03.example.com (192.168.1.113): icmp_seq=1 ttl=64 time=0.386 ms
64 bytes from slave03.example.com (192.168.1.113): icmp_seq=2 ttl=64 time=0.281 ms
^C
--- slave03.example.com ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 1799ms
rtt min/avg/max/mdev = 0.281/0.333/0.386/0.055 ms
测试外网的连通性（我在装的时候，8.8.8.8，已经被禁用….心中一万头cnm）：
[root@slave01 ~]# ping www.baidu.com
ping: unknown host www.baidu.com
[root@slave01 ~]# ping 8.8.8.8
PING 8.8.8.8 (8.8.8.8) 56(84) bytes of data.
From 192.168.1.111 icmp_seq=2 Destination Host Unreachable
From 192.168.1.111 icmp_seq=3 Destination Host Unreachable
From 192.168.1.111 icmp_seq=4 Destination Host Unreachable
From 192.168.1.111 icmp_seq=6 Destination Host Unreachable
From 192.168.1.111 icmp_seq=7 Destination Host Unreachable
From 192.168.1.111 icmp_seq=8 Destination Host Unreachable
（4）SSH完密码登录
#### (1) OpenSSH安装
如果大家在配置时，ping 8.8.8.8能够ping通，则主机能够正常上网；如果不能上网，则将网络连接方式重新设置为NAT，并修改网络配置文件为dhcp方式。在保证网络连通的情况下执行下列命令：
yum install openssh-server- 1
#### (2) 无密码登录实现
使用以下命令生成相应的密钥（三台机器进行相同的操作）
ssh-keygen -t rsa - 1
执行过程一直回车即可
[root@slave01 ~]# ssh-keygen -t rsa
Generating public/private rsa key pair.
Enter file in which to save the key (/root/.ssh/id_rsa):
Enter passphrase (empty for no passphrase):
Enter same passphrase again:
Your identification has been saved in /root/.ssh/id_rsa.
Your public key has been saved in /root/.ssh/id_rsa.pub.
The key fingerprint is:
4e:2f:39:ed:f4:32:2e:a3:55:62:f5:8a:0d:c5:2c:16 [email protected]
The key's randomart image is:
+--[ RSA 2048]----+
| E |
| + |
| o = |
| . + . |
| S . . |
| + X . |
| B * |
| .o=o. |
| .. +oo. |
+-----------------+
生成的文件分别为/root/.ssh/id_rsa（私钥）、/root/.ssh/id_rsa.pub（公钥）
完成后将公钥拷贝到要免登陆的机器上（三台可进行相同操作）：
ssh-copy-id -i slave01.example.com
ssh-copy-id -i slave02.example.com
ssh-copy-id -i slave03.example.com2. Hadoop 2.4.1集群搭建
集群搭建相关软件下载地址：
链接：http://pan.baidu.com/s/1sjIG3b3 密码：38gh下载后将所有软件都放置在E盘的share目录下: 
设置share文件夹为虚拟机的共享目录，如下图所示： 
在linux系统中，采用
[root@slave01 /]# cd /mnt/hgfs/share
[root@slave01 share]# ls命令可以切换到该目录下，如下图 
Spark官方要求的JDK、Scala版本
Spark runs on Java 7+, Python 2.6+ and R 3.1+. For the Scala API, Spark 1.5.0 uses Scala 2.10. You will need to use a compatible Scala version (2.10.x).- 1
（1）JDK 1.8 安装
在根目录下创建sparkLearning目前，后续所有相关软件都放置在该目录下，代码如下：
[root@slave01 /]# mkdir /sparkLearning
[root@slave01 /]# ls
bin etc lib media proc selinux sys var
boot hadoopLearning lib64 mnt root sparkLearning tmp
dev home lost+found opt sbin srv usr
将共享目录中的jdk安装包复制到/sparkLearning目录
[root@slave01 share]# cp /mnt/hgfs/share/jdk-8u40-linux-x64.gz /sparkLearning/
[root@slave01 share]# cd /sparkLearning/
//解压
[root@slave01 sparkLearning]# tar -zxvf jdk-8u40-linux-x64.gz
设置环境变量：
[root@slave01 sparkLearning]# vim /etc/profile在文件最后添加：
export JAVA_HOME=/sparkLearning/jdk1.8.0_40
export PATH=${JAVA_HOME}/bin:$PATH如下图： 
测试配置是否成功：
//使修改后的配置生效
[root@slave01 sparkLearning]# source /etc/profile
//环境变量是否已经设置
[root@slave01 sparkLearning]# $JAVA_HOME
bash: /sparkLearning/jdk1.8.0_40: is a directory
//测试java是否安装配置成功
[root@slave01 sparkLearning]# java -version
java version "1.8.0_40"
Java(TM) SE Runtime Environment (build 1.8.0_40-b25)
Java HotSpot(TM) 64-Bit Server VM (build 25.40-b25, mixed mode)
（2）Scala 2.10.4 安装
//复制文件到sparkLearning目录下
[root@slave01 sparkLearning]# cp /mnt/hgfs/share/scala-2.10.4.tgz .
//解压
[root@slave01 sparkLearning]# tar -zxvf scala-2.10.4.tgz > /dev/null
[root@slave01 sparkLearning]# vim /etc/profile将/etc/profile文件末尾内容修改如下：
export JAVA_HOME=/sparkLearning/jdk1.8.0_40
export SCALA_HOME=/sparkLearning/scala-2.10.4
export PATH=${JAVA_HOME}/bin:${SCALA_HOME}/bin:$PATH测试Scala是否安装成功
[root@slave01 sparkLearning]# source /etc/profile
[root@slave01 sparkLearning]# $SCALA_HOME
bash: /sparkLearning/scala-2.10.4: is a directory
[root@slave01 sparkLearning]# scala -version
Scala code runner version 2.10.4 -- Copyright 2002-2013, LAMP/EPFL
（3）Zookeeper-3.4.5 集群搭建
[root@slave01 sparkLearning]# cp /mnt/hgfs/share/zookeeper-3.4.5.tar.gz .
[root@slave01 sparkLearning]# tar -zxvf zookeeper-3.4.5.tar.gz > /dev/null
[root@slave01 sparkLearning]# cp zookeeper-3.4.5/conf/zoo_sample.cfg zoo.cfg
[root@slave01 sparkLearning]# vim zoo.cfg
修改dataDir为：
dataDir=/sparkLearning/zookeeper-3.4.5/zookeeper_data
在文件末尾添加如下内容：
server.1=slave01.example.com:2888:3888
server.2=slave02.example.com:2888:3888
server.3=slave03.example.com:2888:3888
如图所示： 
创建ZooKeeper集群数据保存目录
[root@slave01 sparkLearning]# cd zookeeper-3.4.5/
[root@slave01 zookeeper-3.4.5]# mkdir zookeeper_data
[root@slave01 zookeeper-3.4.5]# cd zookeeper_data/
[root@slave01 zookeeper_data]# touch myid
[root@slave01 zookeeper_data]# echo 1 > myid
将slave01.example.com（centos_slave01）上的sparkLearning目录拷贝到另外两台服务器上：
[root@slave01 /]# scp -r /sparkLearning slave02.example.com:/
[root@slave01 /]# scp -r /sparkLearning slave03.example.com://etc/profile文件也进行覆盖
[root@slave01 /]# scp /etc/profile slave02.example.com:/etc/profile
[root@slave01 /]# scp /etc/profile slave03.example.com:/etc/profile修改zookeeper_data中的myid信息：
//配置slave02.example.com上的myid
[root@slave01 /]# ssh salve02.example.com
[root@slave02 ~]# echo 2 > /sparkLearning/zookeeper-3.4.5/zookeeper_data/myid
[root@slave02 ~]# more /sparkLearning/zookeeper-3.4.5/zookeeper_data/myid
2
//配置slave03.example.com上的myid
[root@slave02 ~]# ssh slave03.example.com
Last login: Fri Sep 18 01:33:29 2015 from slave01.example.com
[root@slave03 ~]# echo 3 > /sparkLearning/zookeeper-3.4.5/zookeeper_data/myid
[root@slave03 ~]# more /sparkLearning/zookeeper-3.4.5/zookeeper_data/myid
3
如此便完成配置，下面对集群进行测试：
//在slave03.example.com主机上
[root@slave03 ~]# cd /sparkLearning/zookeeper-3.4.5/bin
[root@slave03 bin]# ls
README.txt zkCli.cmd zkEnv.cmd zkServer.cmd
zkCleanup.sh zkCli.sh zkEnv.sh zkServer.sh
//启动slave03.example.com上的ZooKeeper
[root@slave03 bin]# ./zkServer.sh start
JMX enabled by default
Using config: /sparkLearning/zookeeper-3.4.5/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
[root@slave03 bin]# ./zkServer.sh status
JMX enabled by default
Using config: /sparkLearning/zookeeper-3.4.5/bin/../conf/zoo.cfg
Mode: leader
//在slave02.example.com主机上
[root@slave02 bin]# ./zkServer.sh start
JMX enabled by default
Using config: /sparkLearning/zookeeper-3.4.5/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
//查看zookeeper集群状态，如果Mode显示为follower或leader则表明配置成功
[root@slave02 bin]# ./zkServer.sh status
JMX enabled by default
Using config: /sparkLearning/zookeeper-3.4.5/bin/../conf/zoo.cfg
Mode: follower
//在slave01.example.com主机上
[root@slave01 bin]# ./zkServer.sh start
JMX enabled by default
Using config: /sparkLearning/zookeeper-3.4.5/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
[root@slave01 bin]# ./zkServer.sh status
JMX enabled by default
Using config: /sparkLearning/zookeeper-3.4.5/bin/../conf/zoo.cfg
Mode: follower
//在slave03.example.com主机上zookeeper状态
[root@slave03 bin]# ./zkServer.sh status
JMX enabled by default
Using config: /sparkLearning/zookeeper-3.4.5/bin/../conf/zoo.cfg
Mode: leader
（4）Hadoop 2.4.1 集群搭建
（1）Hadoop 2.4.1基本目录浏览
root@slave01 bin]# cp /mnt/hgfs/share/hadoop-2.4.1.tar.gz /sparkLearning/
[root@slave01 bin]# cd /sparkLearning/
[root@slave01 sparkLearning]# tar -zxvf hadoop-2.4.1.tar.gz > /dev/null
[root@slave01 sparkLearning]# cd hadoop-2.4.1
[root@slave01 hadoop-2.4.1]# ls
bin include libexec NOTICE.txt sbin
etc lib LICENSE.txt README.txt share
cd
[root@slave01 hadoop-2.4.1]# cd etc/hadoop/
[root@slave01 hadoop]# ls
capacity-scheduler.xml hdfs-site.xml mapred-site.xml.template
configuration.xsl httpfs-env.sh slaves
container-executor.cfg httpfs-log4j.properties ssl-client.xml.example
core-site.xml httpfs-signature.secret ssl-server.xml.example
hadoop-env.cmd httpfs-site.xml yarn-env.cmd
hadoop-env.sh log4j.properties yarn-env.sh
hadoop-metrics2.properties mapred-env.cmd yarn-site.xml
hadoop-metrics.properties mapred-env.sh
hadoop-policy.xml mapred-queues.xml.template
（2）将Hadoop 2.4.1添加到环境变量
使用命令：vim /etc/profile 将环境变量信息修改如下：
export JAVA_HOME=/sparkLearning/jdk1.8.0_40
export SCALA_HOME=/sparkLearning/scala-2.10.4
export HADOOP_HOME=/sparkLearning/hadoop-2.4.1
export PATH=${JAVA_HOME}/bin:${SCALA_HOME}/bin:${HADOOP_HOME}/bin:${HADOOP_HOME}/sbin:$PATH
（3）将Hadoop 2.4.1添加到环境变量
使用命令：vim hadoop-env.sh 将环境变量信息修改如下，在export JAVA_HOME修改为：
export JAVA_HOME=/sparkLearning/jdk1.8.0_40- 1
（4）修改core-site.xml文件
利用vim core-site.xml命令，文件内容如下：
fs.defaultFS
hdfs://ns1
hadoop.tmp.dir
/sparkLearning/hadoop-2.4.1/tmp
ha.zookeeper.quorum
slave01.example.com:2181,slave02.example.com:2181,slave03.example.com:2181
（5）修改hdfs-site.xml文件
vim hdfs-site.xml内容如下：
dfs.nameservices
ns1
dfs.ha.namenodes.ns1
nn1,nn2
dfs.namenode.rpc-address.ns1.nn1
slave01.example.com:9000
dfs.namenode.http-address.ns1.nn1
slave01.example.com:50070
dfs.namenode.rpc-address.ns1.nn2
slave02.example.com:9000
dfs.namenode.http-address.ns1.nn2
slave02.example.com:50070
dfs.namenode.shared.edits.dir
qjournal://slave01.example.com:8485;slave02.example.com:8485;slave03.example.com:8485/ns1
dfs.journalnode.edits.dir
/sparkLearning/hadoop-2.4.1/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
/home/hadoop/.ssh/id_rsa
dfs.ha.fencing.ssh.connect-timeout
30000
（4）修改mapred-site.xml文件
[root@slave01 hadoop]# cp mapred-site.xml.template mapred-site.xml- 1
vim mapred-site.xml修改文件内容如下:
mapreduce.framework.name
yarn
（6）修改yarn-site.xml文件
yarn.resourcemanager.ha.enabled
true
yarn.resourcemanager.cluster-id
SparkCluster
yarn.resourcemanager.ha.rm-ids
rm1,rm2
yarn.resourcemanager.hostname.rm1
slave01.example.com
yarn.resourcemanager.hostname.rm2
slave02.example.com
yarn.resourcemanager.zk-address
yarn.nodemanager.aux-services
mapreduce_shuffle
（7）修改slaves文件
slave01.example.com
slave02.example.com
slave03.example.com（8）配置文件拷贝到其它服务器
//slave01.example.com上的配置文件拷贝到slave02.example.com
[root@slave01 hadoop]# scp -r /etc/profile slave02.example.com:/etc/profile
profile 100% 2027 2.0KB/s 00:00
[root@slave01 hadoop]# scp -r /sparkLearning/hadoop-2.4.1 slave02.example.com:/sparkLearning/
//slave01.example.com上的配置文件拷贝到slave03.example.com
[root@slave01 hadoop]# scp -r /etc/profile slave03.example.com:/etc/profile
profile 100% 2027 2.0KB/s 00:00
[root@slave01 hadoop]# scp -r /sparkLearning/hadoop-2.4.1 slave03.example.com:/sparkLearning/
（9）启动journalnode
//使用下列命令启动journalnode
[root@slave01 hadoop]# hadoop-daemons.sh start journalnode
slave02.example.com: starting journalnode, logging to /sparkLearning/hadoop-2.4.1/logs/hadoop-root-journalnode-slave02.example.com.out
slave03.example.com: starting journalnode, logging to /sparkLearning/hadoop-2.4.1/logs/hadoop-root-journalnode-slave03.example.com.out
slave01.example.com: starting journalnode, logging to /sparkLearning/hadoop-2.4.1/logs/hadoop-root-journalnode-slave01.example.com.out
//JournalNode进程存在，启动成功
[root@slave01 hadoop]# jps
11261 JournalNode
11295 Jps
[root@slave01 hadoop]# ssh slave02.example.com
Last login: Fri Sep 18 05:33:05 2015 from slave01.example.com
[root@slave02 ~]# jps
6598 JournalNode
6795 Jps
[root@slave02 ~]# ssh slave03.example.com
Last login: Fri Sep 18 05:33:26 2015 from slave02.example.com
[root@slave03 ~]# jps
5876 JournalNode
6047 Jps
[root@slave03 ~]#
（10）格式化HDFS
登录slave02.example.com服务器，执行下列命令
[root@slave02 ~]# hdfs namenode -format
//下面是执行结果
15/09/18 06:05:26 INFO namenode.NameNode: STARTUP_MSG:
/************************************************************
STARTUP_MSG: Starting NameNode
STARTUP_MSG: host = slave02.example.com/127.0.0.1
STARTUP_MSG: args = [-format]
STARTUP_MSG: version = 2.4.1
STARTUP_MSG: classpath = /sparkLearning/hadoop-2.4.1/etc/hadoop:/sparkLearning/hadoop-........省略无关信息...............
STARTUP_MSG: build = http://svn.apache.org/repos/asf/hadoop/common -r 1604318; compiled by 'jenkins' on 2014-06-21T05:43Z
STARTUP_MSG: java = 1.8.0_40
.....................................................省略.....
/sparkLearning/hadoop-2.4.1/tmp/dfs/name has been successfully formatted.
15/09/18 06:05:30 INFO namenode.NNStorageRetentionManager: Going to retain 1 images with txid >= 0
15/09/18 06:05:30 INFO util.ExitUtil: Exiting with status 0
15/09/18 06:05:30 INFO namenode.NameNode: SHUTDOWN_MSG:
/************************************************************
SHUTDOWN_MSG: Shutting down NameNode at slave02.example.com/127.0.0.1
************************************************************/
（11）格式化HDFS信息复制到slave03.example.com服务器
[root@slave02 ~]# scp -r /sparkLearning/hadoop-2.4.1/tmp/ slave01.example.com:/sparkLearning/hadoop-2.4.1/
fsimage_0000000000000000000.md5 100% 62 0.1KB/s 00:00
seen_txid 100% 2 0.0KB/s 00:00
fsimage_0000000000000000000 100% 350 0.3KB/s 00:00
VERSION 100% 200 0.2KB/s 00:00 （12）格式化ZK(在slave02.example.com上执行即可)
[root@slave02 hadoop]# hdfs zkfc -formatZK
Java HotSpot(TM) 64-Bit Server VM warning: You have loaded library /sparkLearning/hadoop-2.4.1/lib/native/libhadoop.so which might have disabled stack guard. The VM will try to fix the stack guard now.
......省略无关信息...............
//执行成功
15/09/18 06:14:22 INFO ha.ActiveStandbyElector: Successfully created /hadoop-ha/ns1 in ZK.
15/09/18 06:14:22 INFO zookeeper.ZooKeeper: Session: 0x34fe096c3ca0000 closed
15/09/18 06:14:22 INFO zookeeper.ClientCnxn: EventThread shut down
（13）启动HDFS(在slave02.example.com上执行)
[root@slave02 hadoop]# start-dfs.sh
[root@slave02 hadoop]# jps
7714 QuorumPeerMain
6598 JournalNode
8295 DataNode
8202 NameNode
8716 Jps
8574 DFSZKFailoverController
[root@slave02 hadoop]# ssh slave01.example.com
Last login: Thu Aug 27 06:24:16 2015 from slave01.example.com
[root@slave01 ~]# jps
13744 DataNode
13681 NameNode
11862 QuorumPeerMain
14007 Jps
13943 DFSZKFailoverController
13851 JournalNode
[root@slave03 ~]# jps
5876 JournalNode
7652 Jps
7068 DataNode
6764 QuorumPeerMain
（14）启动YARN(在slave01.example.com上执行)
//slave01.example.com
[root@slave01 ~]# start-yarn.sh
...输出省略.....
[root@slave01 ~]# jps
14528 Jps
13744 DataNode
13681 NameNode
14228 NodeManager
11862 QuorumPeerMain
13943 DFSZKFailoverController
14138 ResourceManager
13851 JournalNode
//slave02.example.com
[root@slave02 ~]# jps
11216 Jps
10656 JournalNode
7714 QuorumPeerMain
11010 NodeManager
10427 DataNode
10844 DFSZKFailoverController
10334 NameNode
//slave03.example.com
[root@slave03 ~]# jps
8610 JournalNode
8791 NodeManager
8503 DataNode
9001 Jps
6764 QuorumPeerMain
（15）查看hadoop运行管理界面
打开浏览器，输入http://slave01.example.com:8088/，可以得到hadoop集群管理界面: 
输入http://slave01.example.com:50070 可以得到HDFS管理界面 
至此Hadoop集群配置成功
3. Spark 1.5.0 集群部署
（1）将Spark添加到环境变量
[root@slave01 hadoop]# cp /mnt/hgfs/share/spark-1.5.0-bin-hadoop2.4.tgz /sparkLearning/
[root@slave01 sparkLearning]# tar -zxvf spark-1.5.0-bin-hadoop2.4.tgz > /dev/null
[root@slave01 sparkLearning]# vim /etc/profile将/etc/profile内容修改如下：
export JAVA_HOME=/sparkLearning/jdk1.8.0_40
export SCALA_HOME=/sparkLearning/scala-2.10.4
export HADOOP_HOME=/sparkLearning/hadoop-2.4.1
export SPARK_HOME=/sparkLearning/spark-1.5.0-bin-hadoop2.4
export PATH=${JAVA_HOME}/bin:${SCALA_HOME}/bin:${HADOOP_HOME}/bin:${HADOOP_HOME}/sbin:${SPARK_HOME}/bin:${SPARK_HOME}/sbin:$PATH
（2）将Spark添加到环境变量
[root@slave01 sparkLearning]# cd spark-1.5.0-bin-hadoop2.4/conf
[root@slave01 conf]# ls
docker.properties.template metrics.properties.template spark-env.sh.template
fairscheduler.xml.template slaves.template
log4j.properties.template spark-defaults.conf.template
//复制模板文件
[root@slave01 conf]# cp spark-env.sh.template spark-env.sh
[root@slave01 conf]# vim spark-env.sh在spark-env.sh文件中添加如下内容：
export JAVA_HOME=/sparkLearning/jdk1.8.0_40
export SCALA_HOME=/sparkLearning/scala-2.10.4
export HADOOP_CONF_DIR=/sparkLearning/hadoop-2.4.1/etc/hadoop[root@slave01 conf]# cp slaves.template slaves
[root@slave01 conf]# vim slaves
slaves文件内容如下：
# A Spark Worker will be started on each of the machines listed below.
slave01.example.com
slave02.example.com
slave03.example.com（3）将配置信息复制到其它服务器
[root@slave01 sparkLearning]# scp /etc/profile slave02.example.com:/etc/profile
profile 100% 2123 2.1KB/s 00:00
[root@slave01 sparkLearning]# scp /etc/profile slave03.example.com:/etc/profile
profile 100% 2123 2.1KB/s 00:00
[root@slave01 sparkLearning]# vim /etc/profile
[root@slave01 sparkLearning]# scp -r spark-1.5.0-bin-hadoop2.4 slave02.example.com:/sparkLearning/
...执行过程省略.....
[root@slave01 sparkLearning]# scp -r spark-1.5.0-bin-hadoop2.4 slave03.example.com:/sparkLearning/
...执行过程省略.....
（4）启动Spark集群
因为本人机器上装了Ambari Server，占用了8080端口，而Spark Master默认端是8080，因此将sbin/start-master.sh中的SPARK_MASTER_WEBUI_PORT修改为8888
if [ "$SPARK_MASTER_WEBUI_PORT" = "" ]; then
SPARK_MASTER_WEBUI_PORT=8888
fi[root@slave01 sbin]# ./start-all.sh
starting org.apache.spark.deploy.master.Master, logging to /sparkLearning/spark-1.5.0-bin-hadoop2.4/sbin/../logs/spark-root-org.apache.spark.deploy.master.Master-1-slave01.example.com.out
slave03.example.com: starting org.apache.spark.deploy.worker.Worker, logging to /sparkLearning/spark-1.5.0-bin-hadoop2.4/sbin/../logs/spark-root-org.apache.spark.deploy.worker.Worker-1-slave03.example.com.out
slave02.example.com: starting org.apache.spark.deploy.worker.Worker, logging to /sparkLearning/spark-1.5.0-bin-hadoop2.4/sbin/../logs/spark-root-org.apache.spark.deploy.worker.Worker-1-slave02.example.com.out
slave01.example.com: starting org.apache.spark.deploy.worker.Worker, logging to /sparkLearning/spark-1.5.0-bin-hadoop2.4/sbin/../logs/spark-root-org.apache.spark.deploy.worker.Worker-1-slave01.example.com.out
[root@slave01 sbin]# jps
13744 DataNode
13681 NameNode
14228 NodeManager
16949 Master
11862 QuorumPeerMain
13943 DFSZKFailoverController
14138 ResourceManager
13851 JournalNode
17179 Jps
17087 Worker
浏览器中输入slave01.example.com:8888 
但是在启动过程中出现了错误，查看日志文件
[root@slave02 logs]# more spark-root-org.apache.spark.deploy.worker.Worker-1-slave02.example.com.out
日志内容中包括下列错误：
akka.actor.ActorNotFound: Actor not found for: ActorSelection[Anchor(akka.tcp://
[email protected]:7077/), Path(/user/Master)]
at akka.actor.ActorSelection$$anonfun$resolveOne$1.apply(ActorSelection.
scala:65)
at akka.actor.ActorSelection$$anonfun$resolveOne$1.apply(ActorSelection.
scala:63)
at scala.concurrent.impl.CallbackRunnable.run(Promise.scala:32)
at akka.dispatch.BatchingExecutor$AbstractBatch.processBatch(BatchingExe
cutor.scala:55)
at akka.dispatch.BatchingExecutor$Batch.run(BatchingExecutor.scala:73)
at akka.dispatch.ExecutionContexts$sameThreadExecutionContext$.unbatched
Execute(Future.scala:74)
at akka.dispatch.BatchingExecutor$class.execute(BatchingExecutor.scala:1
20)
at akka.dispatch.ExecutionContexts$sameThreadExecutionContext$.execute(F
uture.scala:73)
at scala.concurrent.impl.CallbackRunnable.executeWithValue(Promise.scala
:40)
at scala.concurrent.impl.Promise$DefaultPromise.tryComplete(Promise.scal
a:248)
at akka.pattern.PromiseActorRef.$bang(AskSupport.scala:266)
at akka.actor.EmptyLocalActorRef.specialHandle(ActorRef.scala:533)
at akka.actor.DeadLetterActorRef.specialHandle(ActorRef.scala:569)
.....省略.....................没找到具体原因，在ubuntu 10.04服务器上进行相同的配置，集群搭建却成功了（心中一万头…..），运行界面如下: 
（5）测试Spark集群
采用下列命上传spark-1.5.0-bin-hadoop2.4目录下的README.md文件到相应的根目录。
hadoop dfs -put README.md 如下图： 
进入/spark-1.5.0-bin-hadoop2.4/bin目录，启动./spark-shell，如下图所示： 
执行REDME.md文件的wordcount操作：
scala> val textCount = sc.textFile(“README.md”).filter(line => line.contains(“Spark”)).count()
如下图： 
执行结果如下图： 
至此，Spark 1.5集群搭建成功。