Spark安装过程

Precondition：jdk、Scala安装，/etc/profile文件部分内容如下：

JAVA_HOME=/home/Spark/husor/jdk
CLASSPATH=.:$JAVA_HOME/jre/lib/rt.jar:$JAVA_HOME/lib/dt.jar:$JAVA_HOME/lib/tools.jar
export JAVA_HOME
export CLASSPATH

HADOOP_HOME=/home/Spark/husor/hadoop
HBASE_HOME=/home/Spark/husor/hbase
SCALA_HOME=/home/Spark/husor/scala
SPARK_HOME=/home/Spark/husor/spark
PATH=$JAVA_HOME/bin:$HADOOP_HOME/bin:$HBASE_HOME/bin:$SCALA_HOME/bin:$SPARK_HOME/bin:$PATH
export HADOOP_HOME
export HBASE_HOME
export SCALA_HOME
export SPARK_HOME
"/etc/profile" 99L, 2415C written
[root@Master husor]# source /etc/profile
[root@Master husor]# echo $SPARK_HOME
/home/Spark/husor/spark
[root@Master husor]# echo $SCALA_HOME
/home/Spark/husor/scala
[root@Master husor]# scala -version
Scala code runner version 2.10.4 -- Copyright 2002-2013, LAMP/EPFL

　　

1. expect安装

Expect是基于Tcl语言的一种脚本语言，其实无论是交互还是非交互的应用场合，Expect都可以大显身手，但是对于交互式的特定场合，还非Except莫属。

第1步：使用root用户登录 
    
第2步：下载安装文件expect-5.43.0.tar.gz 和 tcl8.4.11-src.tar.gz 
       
第3步：解压安装包 
       解压tcl8.4.11-src.tar.gz 
       tar –xvf tcl8.4.11-src.tar.gz 
       解压后将创建tcl8.4.11 文件夹

解压expect-5.43.0.tar.gz 
       tar –xvf expect-5.43.0.tar.gz 
       解压后将创建expect-5.43 文件夹 
       
第4步：安装tcl 
       进入tcl8.4.11/unix 目录 
        a.执行sed -i "s/relid'/relid/" configure 
        b.执行./configure --prefix=/expect 
        c.执行make 
        d.执行make install 
        e.执行mkdir -p /tools/lib 
        f.执行cp tclConfig.sh /tools/lib/ 
        g. 将/tools/bin目录export到环境变量 
           tclpath=/tools/bin 
           export tclpath

第5步：安装Expect 
        进入/soft/expect-5.43目录 
        执行./configure --prefix=/tools --with-tcl=/tools/lib --with-x=no 
        如果最后一行提示： 
        configure: error: Can't find Tcl private headers 
        需要添加一个头文件目录参数 
        --with-tclinclude=../tcl8.4.11/generic,即 
        ./configure --prefix=/tools --with-tcl=/tools/lib --with-x=no --with-tclinclude=../tcl8.4.11/generic 
        ../tcl8.4.11/generic 就是tcl解压安装后的路径，一定确保该路径存在 
        执行make 
        执行make install 
        编译完成后会生在/tools/bin内生成expect命令 
        执行/tools/bin/expect出现expect1.1>提示符说明expect安装成功.

第6步：创建一个符号链接 
        ln -s /tools/bin/expect /usr/bin/expect 
        查看符号连接 
        ls -l /usr/bin/expect 
        lrwxrwxrwx 1 root root 17 06-09 11:38 /usr/bin/expect -> /tools/bin/expect

这个符号链接将在编写expect脚本文件时用到，例如在expect文件头部会指定用于执行该脚本的shell 
        #!/usr/bin/expect

2. SSH免输入密码登陆

主机Master操作如下：

[Spark@Master ~]$ ssh-keygen -t rsa -P '' -f ~/.ssh/id_rsa
Generating public/private rsa key pair.
Your identification has been saved in /home/Spark/.ssh/id_rsa.
Your public key has been saved in /home/Spark/.ssh/id_rsa.pub.
The key fingerprint is:
c9:d0:1f:92:43:42:85:f1:c5:23:76:f8:df:80:e5:66 Spark@Master
The key's randomart image is:
+--[ RSA 2048]----+
| .++oo. |
| .=+o+ . |
| ..*+.= |
| o =o.E |
| S .+ o |
| . . |
| |
| |
| |
+-----------------+
[Spark@Master ~]$ cat ~/.ssh/id_rsa.pub >> ~/.ssh/authorized_keys

3. 然后执行如下自动化传输公钥脚本SSH.sh，将主机Master上的公钥传输给各个从节点Slave1，Slave2......

（Note：将SSH.sh和NoPwdAccessSSH.exp脚本文件添加执行权限，如下：）

[Spark@Master test]$ chmod +x SSH.sh

[Spark@Master test]$ chmod +x NoPwdAccessSSH.exp

//执行自动化无密码访问脚本SSH.sh

[Spark@Master test]$ ./SSH.sh
spawn ssh-copy-id -i /home/Spark/.ssh/id_rsa.pub Spark@Master
The authenticity of host 'master (192.168.8.29)' can't be established.
RSA key fingerprint is f0:3f:04:51:36:b5:91:c7:fa:47:5a:49:bc:fd:fe:40.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added 'master,192.168.8.29' (RSA) to the list of known hosts.
Now try logging into the machine, with "ssh 'Spark@Master'", and check in:

.ssh/authorized_keys

to make sure we haven't added extra keys that you weren't expecting.

No Password Access Master is Succeed!!!
spawn ssh-copy-id -i /home/Spark/.ssh/id_rsa.pub Spark@Slave1
Spark@slave1's password:
Now try logging into the machine, with "ssh 'Spark@Slave1'", and check in:

.ssh/authorized_keys

to make sure we haven't added extra keys that you weren't expecting.

No Password Access Slave1 is Succeed!!!
spawn ssh-copy-id -i /home/Spark/.ssh/id_rsa.pub Spark@Slave2
Spark@slave2's password:
Now try logging into the machine, with "ssh 'Spark@Slave2'", and check in:

.ssh/authorized_keys

to make sure we haven't added extra keys that you weren't expecting.

No Password Access Slave2 is Succeed!!!
[Spark@Master test]$ ssh Slave1
Last login: Wed Nov 19 02:35:28 2014 from 192.168.8.29
Welcome to your pre-built HUSOR STANDARD WEB DEVELOP VM.

PHP5.3 (/usr/local/php-cgi) service:php-fpm
PHP5.4 (/usr/local/php-54) service:php54-fpm
Tengine1.4.6, mysql-5.5.29, memcached 1.4.15, tokyocabinet-1.4.48, tokyotyrant-1.1.41, httpsqs-1.7, coreseek-4.1

WEBROOT: /data/webroot/www/

[Spark@Slave1 ~]$ exit
logout
Connection to Slave1 closed.
[Spark@Master test]$ ssh Slave2
Last login: Wed Nov 19 01:48:01 2014 from 192.168.8.1
Welcome to your pre-built HUSOR STANDARD WEB DEVELOP VM.

PHP5.3 (/usr/local/php-cgi) service:php-fpm
PHP5.4 (/usr/local/php-54) service:php54-fpm
Tengine1.4.6, mysql-5.5.29, memcached 1.4.15, tokyocabinet-1.4.48, tokyotyrant-1.1.41, httpsqs-1.7, coreseek-4.1

WEBROOT: /data/webroot/www/

[Spark@Slave2 ~]$

以上自动化执行脚本文件如下：

SSH.sh

#!/bin/bash

bin=`which $0`

bin=`dirname ${bin}`
bin=`cd "$bin"; pwd`

if [ ! -x "$bin/NoPwdAccessSSH.exp" ]; then
  echo "Sorry, $bin/NoPwdAccessSSH.exp is not executable file,please chmod +x $bin/NoPwdAccessSSH.exp."
  exit 1
fi

for hostInfo in $(cat $bin/SparkCluster);do

    host_name=$(echo "$hostInfo"|cut -f1 -d":")
    user_name=$(echo "$hostInfo"|cut -f2 -d":")
    user_pwd=$(echo "$hostInfo"|cut -f3 -d":")

    local_host=`ifconfig eth0 | grep "Mask" | cut -d: -f2 | awk '{print $1}'`
    if [ $host_name = $local_host ]; then
        continue;
    else
        expect $bin/NoPwdAccessSSH.exp $host_name $user_name $user_pwd //调用expect应答式脚本NoPwdAccessSSH.exp
    fi

    if [ $? -eq 0 ]
    then
        echo "No Password Access $host_name is Succeed!!!"
    else
        echo "No Password Access $host_name is failed!!!"
    fi

done

NoPwdAccessSSH.exp

#!/usr/bin/expect -f

# auto ssh login

if { $argc<3} {
  puts stderr "Usage: $argv0(hostname) $argv1(username) $argv2(userpwd).\n "
　exit 1
}

set hostname [lindex $argv 0]
set username [lindex $argv 1]
set userpwd [lindex $argv 2]

spawn ssh-copy-id -i /home/Spark/.ssh/id_rsa.pub $username@$hostname

expect {
   "*yes/no*" { send "yes\r";exp_continue }
   "*password*" { send "$userpwd\r";exp_continue }
   "*password*" { send "$userpwd\r"; }
}

其中的SparkCluster文件内容如下：

Master:Spark:111111
Slave1:Spark:111111
Slave2:Spark:111111

3. 安装hadoop2.4.1(呵呵，我博客上有的。。。。。。)

Note：

1> 将hadoop，jdk安装到统一新添用户Spark相应目录下：/home/Spark）（不然会引起一系列权限问题）

2> 将hadoop安装目录bin和sbin下添加执行权限(chmod 777 *)

3> 将主机Master上配置好的hadoop安装目录scp到所有从机Slave相同的新增用户Spark相同目录下：(/home/Spark) -> scp -r /home/Spark/* Spark@SlaveX:/home/Spark

4> 统一使用root用户修改/etc/hosts，添加相关hostname识别(192.168.8.29 Master 192.168.8.30 Slave1 192.168.8.31 Slave2)

所遇异常1：

Hadoop 2.2.0 - warning: You have loaded library /home/hadoop/2.2.0/lib/native/libhadoop.so.1.0.0 which might have disabled stack guard.

Unable to load native-hadoop library for your platform... using builtin-java classes where applicable

Starting namenodes on [Java HotSpot(TM) 64-Bit Server VM warning: You have loaded library /home/Spark/hadoop2.4.1/lib/native/libhadoop.so.1.0.0 which might have disabled stack guard. The VM will try to fix the stack guard now.

It's highly recommended that you fix the library with 'execstack -c <libfile>', or link it with '-z noexecstack'.

localhost]

sed: -e expression #1, char 6: unknown option to `s'

HotSpot(TM): ssh: Could not resolve hostname HotSpot(TM): Name or service not known

64-Bit: ssh: Could not resolve hostname 64-Bit: Name or service not known

Java: ssh: Could not resolve hostname Java: Name or service not known

Server: ssh: Could not resolve hostname Server: Name or service not known

VM: ssh: Could not resolve hostname VM: Name or service not known

 

Reason：

因为官网下载的prebuild hadoop中使用的本地库文件(例如lib/native/libhadoop.so.1.0.0)都是基于32位编译的，运行在64位系统上就会出现上述错误。

 

解决方案1：

在64位系统上重新编译hadoop

 

解决方案2：

以root用户在/etc/profile中添加：

export HADOOP_COMMON_LIB_NATIVE_DIR=${HADOOP_HOME}/lib/native
export HADOOP_OPTS="-Djava.library.path=$HADOOP_HOME/lib"

使对应配置立即生效：

source /etc/profile

 

解决方案3：

在hadoop-env.sh和yarn-env.sh中添加如下两行：

export HADOOP_HOME=/home/Spark/hadoop

export HADOOP_COMMON_LIB_NATIVE_DIR=${HADOOP_HOME}/lib/native
export HADOOP_OPTS="-Djava.library.path=$HADOOP_HOME/lib"

 

namenode格式化:

bin/hdfs namenode -format

启动或停止namenode，datanode：

sbin/start-dfs.sh    -> sbin/stop-dfs.sh

启动或停止resourcemanage和nodemanager资源管理器

sbin/start-yarn.sh  -> sbin/stop-yarn.sh

 

所遇异常2：

当用window 7浏览器查看Hadoop相关界面时，即http://Master:50070时，window 7无法识别Master，当使用Master对应的IP地址时，是可以查看相关界面的。

Reason：

window 7无法识别Master。

解决方案：

在win7地址栏上输入 %systemroot%\system32\drivers\etc 内容,回车后就能看到 hosts 文件了，添加对应主机名识别即可(192.168.8.29 Master 192.168.8.30 Slave1 192.168.8.31 Slave2)。

 

4. 验证界面

 

 

5. Spark集群安装

 

配置spark-env.sh文件

添加如下内容：

export JAVA_HOME=/home/Spark/husor/jdk
export HADOOP_HOME=/home/Spark/husor/hadoop
export HADOOP_CONF_DIR=/home/Spark/husor/hadoop/etc/hadoop
export SCALA_HOME=/home/Spark/husor/scala
export SPARK_MASTER_IP=Master
export SPARK_WORKER_MEMORY=512m

配置slaves文件

删除localhost，添加相关内容：

Slave1

Slave2

验证Spark启动

 

Spark Shell启动

[Spark@Master spark]$ bin/spark-shell
Spark assembly has been built with Hive, including Datanucleus jars on classpath
// :: INFO spark.SecurityManager: Changing view acls to: Spark,
// :: INFO spark.SecurityManager: Changing modify acls to: Spark,
// :: INFO spark.SecurityManager: SecurityManager: authentication disabled; ui acls disabled; users with view permissions: Set(Spark, ); users with modify permissions: Set(Spark, )
// :: INFO spark.HttpServer: Starting HTTP Server
// :: INFO server.Server: jetty-.y.z-SNAPSHOT
// :: INFO server.AbstractConnector: Started SocketConnector@0.0.0.0:
// :: INFO util.Utils: Successfully started service 'HTTP class server' on port .
Welcome to
      ____              __
     / __/__  ___ _____/ /__
    _\ \/ _ \/ _ `/ __/  '_/
   /___/ .__/\_,_/_/ /_/\_\   version 1.1.
      /_/

Using Scala version 2.10. (Java HotSpot(TM) -Bit Server VM, Java 1.7.0_71)
Type in expressions to have them evaluated.
Type :help for more information.
// :: INFO spark.SecurityManager: Changing view acls to: Spark,
// :: INFO spark.SecurityManager: Changing modify acls to: Spark,
// :: INFO spark.SecurityManager: SecurityManager: authentication disabled; ui acls disabled; users with view permissions: Set(Spark, ); users with modify permissions: Set(Spark, )
// :: INFO slf4j.Slf4jLogger: Slf4jLogger started
// :: INFO Remoting: Starting remoting
// :: INFO Remoting: Remoting started; listening on addresses :[akka.tcp://sparkDriver@Master:38507]
// :: INFO Remoting: Remoting now listens on addresses: [akka.tcp://sparkDriver@Master:38507]
// :: INFO util.Utils: Successfully started service 'sparkDriver' on port .
// :: INFO spark.SparkEnv: Registering MapOutputTracker
// :: INFO spark.SparkEnv: Registering BlockManagerMaster
// :: INFO storage.DiskBlockManager: Created local directory at /tmp/spark-local--651a
// :: INFO util.Utils: Successfully started service 'Connection manager for block manager' on port .
// :: INFO network.ConnectionManager: Bound socket to port  with id = ConnectionManagerId(Master,)
// :: INFO storage.MemoryStore: MemoryStore started with capacity 267.3 MB
// :: INFO storage.BlockManagerMaster: Trying to register BlockManager
// :: INFO storage.BlockManagerMasterActor: Registering block manager Master: with 267.3 MB RAM
// :: INFO storage.BlockManagerMaster: Registered BlockManager
// :: INFO spark.HttpFileServer: HTTP File server directory is /tmp/spark-7decc3d6-acce--98c3-172c680de719
// :: INFO spark.HttpServer: Starting HTTP Server
// :: INFO server.Server: jetty-.y.z-SNAPSHOT
// :: INFO server.AbstractConnector: Started SocketConnector@0.0.0.0:
// :: INFO util.Utils: Successfully started service 'HTTP file server' on port .
// :: INFO server.Server: jetty-.y.z-SNAPSHOT
// :: INFO server.AbstractConnector: Started SelectChannelConnector@0.0.0.0:
// :: INFO util.Utils: Successfully started service 'SparkUI' on port .
// :: INFO ui.SparkUI: Started SparkUI at http://Master:4040
// :: WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
// :: INFO executor.Executor: Using REPL class URI: http://192.168.8.29:34246
// :: INFO util.AkkaUtils: Connecting to HeartbeatReceiver: akka.tcp://sparkDriver@Master:38507/user/HeartbeatReceiver
// :: INFO repl.SparkILoop: Created spark context..
Spark context available as sc.

scala>