完全从0搭建Spark集群
备注：这个步骤，只适合用root来搭建，正式环境下应该要有权限类的东西后面另外再进行实验写教程
1、安装各个软件，设置环境变量（每种软件需自己单独下载）
export JAVA_HOME=/usr/java/jdk1.8.0_71
export JAVA_BIN=/usr/java/jdk1.8.0_71/bin
export PATH=$JAVA_HOME/bin:$PATH
export CLASSPATH=.:$JAVA_HOME/lib/dt.jar:$JAVA_HOME/lib/tools.jar
export JAVA_HOME JAVA_BIN PATH CLASSPATH
export HADOOP_HOME=/usr/local/hadoop-2.6.0
export HADOOP_CONF_DIR=${HADOOP_HOME}/etc/hadoop
export HADOOP_COMMON_LIB_NATIVE_DIR=${HADOOP_HOME}/lib/native
export HADOOP_OPTS="-Djava.library.path=${HADOOP_HOME}/lib"
export PATH=${HADOOP_HOME}/bin:${HADOOP_HOME}/sbin:$PATH
export SCALA_HOME=/usr/local/scala-2.10.4
export PATH=${SCALA_HOME}/bin:$PATH
export SPARK_HOME=/usr/local/spark/spark-1.6.0-bin-hadoop2.6
export PATH=${SPARK_HOME}/bin:${SPARK_HOME}/sbin:$PATH
export ZOOKEEPER_HOME=/usr/local/zookeeper-3.4.6
2、ssh设置
ssh-keygen -t dsa -P ‘‘ -f ~/.ssh/id_dsa //生成key到~/.ssh/id_dsa中
cat ~/.ssh/id_dsa.pub >> ~/.ssh/authorized_keys //追加到key中
3、主机名和域名的设置
vi /etc/hostname 改成Master或者Worker1、2、3、4
vim /etc/hosts 改域名，各个系统ip对应的域名
4、Hadoop的配置
1）cd $HADOOP_HOME/etc/hadoop/ 下改动core-site.xml
<configuration>
<property>
<name>fs.defaultFS</name>
<value>hdfs://Master:9000</value>
</property>
<property>
<name>hadoop.tmp.dir</name>
<value>/usr/local/hadoop/hadoop-2.6.0/tmp</value>
</property>
<property>
<name>hadoop.native.lib</name>
<value>true</value>
<description>Should native hadoop libraries,if present,be used</description>
</property>
</configuration>
2）还是cd $HADOOP_HOME/etc/hadoop/ 下改动hdfs-site.xml
<configuration>
<property>
<name>dfs.replication</name>
<value>2</value>
</property>
<property>
<name>dfs.namenode.secondary.http-address</name>
<value>Master:50090</value>
<description>The secondary namenode http server address and port</description>
</property>
<property>
<name>dfs.namenode.name.dir</name>
<value>/usr/local/hadoop/hadoop-2.6.0/dfs/name</value>
</property>
<property>
<name>dfs.datanode.dir</name>
<value>/usr/local/hadoop/hadoop-2.6.0/dfs/data</value>
</property>
<property>
<name>dfs.namenode.checkpoint.dir</name>
<value>file:///usr/local/hadoop/hadoop-2.6.0/dfs/namesecondary</value>
<description>Determines where on the local filesystem the DFSsecondary name node should store th temporary images to merge.If this is acomma-delimited list of directories then the image is replicated in all of the irectories foe redundancy.</description>
</property>
</configuration>
3）还是cd $HADOOP_HOME/etc/hadoop/ 下改动 mapred-site.xml
<configuration>
<property>
<name>mapreduce.framework.name</name>
<value>yarn</value>
</property>
</configuration>
4）还是cd $HADOOP_HOME/etc/hadoop/ 下改动 yarn-site.xml
<configuration>
<property>
<name>yarn.resourcemanager.hostname</name>
<value>Master</value>
</property>
<property>
<name>yarn.nodemanager.aux-services</name>
<value>mapreduce_shuffle</value>
</property>
</configuration>
5）还是cd $HADOOP_HOME/etc/hadoop/ 下改动 hadoop-env.sh
export JAVA_HOME=/usr/java/jdk1.8.0_71 对应的jdk目录
如果你想把你的Master也作为一个节点，也可以Master加入，但是机器不够多Driver如果是Master，Driver还有其它程序比如web查询等在跑，不建议把Master作为节点来跑。
=======一台机器配置到上面这个程度，然后开始复制机器，再继续下面的操作=============
6）先再看第3步里面改域名的，到各个机器上把主机名称和域名改好了
7）还是cd $HADOOP_HOME/etc/hadoop/ 下改动 slaves
看你有多少slaves，分别把那些机器的域名加入进来，比如
Worker1
Worker2
Worker3
然后分别复制给几台机器：
scp slaves root@Worker1:/usr/local/hadoop-2.6.0/etc/hadoop/slaves
scp slaves root@Worker2:/usr/local/hadoop-2.6.0/etc/hadoop/slaves
scp slaves root@Worker3/usr/local/hadoop-2.6.0/etc/hadoop/slaves
8）还是cd $HADOOP_HOME/etc/hadoop/ 下改动 Master，内容就是Master
在Master不做集群的情况下，需要把Master分别拷贝到各个机器上，其实应该拷贝，这样如果不启动集群，也能跑
如果Master是集群，ZooKeeper，在ZooKeeper进行配置
scp Master root@Worker1:/usr/local/hadoop-2.6.0/etc/hadoop/Master
scp Master root@Worker2:/usr/local/hadoop-2.6.0/etc/hadoop/Master
scp Master root@Worker3/usr/local/hadoop-2.6.0/etc/hadoop/Master
9）在Master上格式化系统
mkdir /usr/local/hadoop/hadoop-2.6.0/tmp 如果原来存在，就删除
hdfs namenode -format
10)启动dfs
cd $HADOOP_HOME/sbin
./start-dfs.sh
然后
http://Master:50070/dfshealth.html 可以看dfs文件状态
看不到，比如Configured Capacity只有0B，尝试每台机器防火墙关闭：
systemctl stop firewalld.service
systemctl disable firewalld.service
但是这个只适合于开发机，实际生产环境需要仔细看什么端口来确定。
***************如果只做Spark，到这里就足够了，做Hadoop另外说****************
5、Spark的配置
1）spark-env.sh
cd $SPARK_HOME/conf
cp出来spark-env.sh
export JAVA_HOME=/usr/java/jdk1.8.0_71
export SCALA_HOME=/usr/local/scala-2.10.4
export HADOOP_HOME=/usr/local/hadoop-2.6.0
export HADOOP_CONF_DIR=/usr/local/hadoop-2.6.0/etc/hadoop
#export SPARK_CLASSPATH=$SPARK_CLASSPATH:$SPARK_HOME/lib/ojdbc-14.jar:$SPARK_HOME/lib/jieyi-tools-1.2.0.7.RELEASE.jar
#export SPARK_MASTER_IP=Master
export SPARK_WORKER_MEMORY=2g
export SPARK_EXCUTOR_MEMORY=2g
export SPARK_DRIVER_MEMORY=2g
export SPARK_WORKER_CORES=8
export SPARK_DAEMON_JAVA_OPTS="-Dspark.deploy.recoveryMode=ZOOKEEPER -Dspark.deploy.zookeeper.url=Master:2181,Worker1:2181,Worker2:2181 -Dspark.deploy.zookeeper.dir=/spark"
参数意思讲解：
export JAVA_HOME=/usr/java/jdk1.8.0_71
export SCALA_HOME=/usr/local/scala-2.10.4
export HADOOP_HOME=/usr/local/hadoop-2.6.0
export HADOOP_CONF_DIR=/usr/local/hadoop-2.6.0/etc/hadoop //运行在yarn模式下必须配置
export SPARK_MASTER_IP=Master //Saprk运行的主ip
export SPARK_WORKER_MEMORY=2g //具体机器
export SPARK_EXCUTOR_MEMORY=2g //具体计算
export SPARK_DRIVER_MEMORY=2g
export SPARK_WORKER_CORES=8 //线程池并发数
其中export SPARK_MASTER_IP=Master是作为单机的时候，export SPARK_DAEMON_JAVA_OPTS="-Dspark.deploy.recoveryMode=ZOOKEEPER -Dspark.deploy.zookeeper.url=Master:2181,Worker1:2181,Worker2:2181 -Dspark.deploy.zookeeper.dir=/spark"是作为集群的时候的配置
改完之后同步：
scp spark-env.sh root@Worker1:/usr/local/spark/spark-1.6.0-bin-hadoop2.6/conf/spark-env.sh
scp spark-env.sh root@Worker2:/usr/local/spark/spark-1.6.0-bin-hadoop2.6/conf/spark-env.sh
scp spark-env.sh root@Worker3:/usr/local/spark/spark-1.6.0-bin-hadoop2.6/conf/spark-env.sh
2）slaves
cd $SPARK_HOME/conf
cp出来slaves
内容为为：
Worker1
Worker2
Worker3
改完之后同步：
scp slaves root@Worker1:/usr/local/spark/spark-1.6.0-bin-hadoop2.6/conf/slaves
scp slaves root@Worker2:/usr/local/spark/spark-1.6.0-bin-hadoop2.6/conf/slaves
scp slaves root@Worker3:/usr/local/spark/spark-1.6.0-bin-hadoop2.6/conf/slaves
3）spark-defaults.conf
cd $SPARK_HOME/conf
cp spark-defaults.conf出来
spark.executor.extraJavaOptions -XX:+PrintGCDetails -Dkey=value -Dnumbers="one two three"
spark.eventLog.enabled true
spark.eventLog.dir hdfs://Master:9000/historyserverforSpark1
spark.yarn.historyServer.address Master:18080
spark.history.fs.logDirectory hdfs://Master:9000/historyserverforSpark1
改完之后同步：
scp spark-defaults.conf root@Worker1:/usr/local/spark/spark-1.6.0-bin-hadoop2.6/conf/spark-defaults.conf
scp spark-defaults.conf root@Worker2:/usr/local/spark/spark-1.6.0-bin-hadoop2.6/conf/spark-defaults.conf
scp spark-defaults.conf root@Worker3:/usr/local/spark/spark-1.6.0-bin-hadoop2.6/conf/spark-defaults.conf
或者上面三步一起：
cd $SPARK_HOME
scp -r ./spark-1.6.0-bin-hadoop2.6/ root@Worker1:/usr/local/spark
4）创建历史目录（第一次安装必须做）
hadoop dfs -rmr /historyserverforSpark
hadoop dfs -mkdir /historyserverforSpark
然后这里就有了：
5）启动spark
cd $SPARK_HOME/sbin
./start-all.sh
看web控制台
master:8080/
6）启动历史信息的服务
cd $SPARK_HOME/sbin
./start-history-server.sh
7）实验下Pi的算法：
./spark-submit --class org.apache.spark.examples.SparkPi --master spark://Master:7077 ../lib/spark-examples-1.6.0-hadoop2.6.0.jar 100
开始神奇的Spark之旅吧！
***************如果Spark单机，到这里就足够了，开始补充Zookeeper做集群的东西****************
6、ZooKeeper安装集群的东西
1）先在第一台机器上解压zookeeper，目录按照开头的环境变量解压就可以
进入到zookeeper下，创建data和logs两个目录
root@Master:/usr/local/zookeeper-3.4.6# mkdir data
root@Master:/usr/local/zookeeper-3.4.6# mkdir logs
2）从zoo_sample.cfg中cp出zoo.cfg并设置
root@Master:/usr/local/zookeeper-3.4.6/conf# cp zoo_sample.cfg zoo.cfg
root@Master:/usr/local/zookeeper-3.4.6/conf# vi zoo.cfg
修改(做3台机器的集群)
dataDir=/usr/local/zookeeper-3.4.6/data
dataLogDir=/usr/local/zookeeper-3.4.6/logs
server.0=Master:2888:3888
server.1=Worker1:2888:3888
server.2=Worker2:2888:3888
3）在data下面为机器编号
root@Master:/usr/local/zookeeper-3.4.6/conf# cd ../data/
为机器编号
root@Master:/usr/local/zookeeper-3.4.6/data# echo 0>myid
root@Master:/usr/local/zookeeper-3.4.6/data# echo 0>>myid
root@Master:/usr/local/zookeeper-3.4.6/data# ls
myid
root@Master:/usr/local/zookeeper-3.4.6/data# cat myid
root@Master:/usr/local/zookeeper-3.4.6/data# vi myid 在这里在里面写一个0
root@Master:/usr/local/zookeeper-3.4.6/data# cat myid
0
到这个时候一台机器已经配置好了
4）拷贝给其它两台机器同时更改myid
root@Master:/usr/local# scp -r ./zookeeper-3.4.6 root@Worker1:/usr/local
root@Master:/usr/local# scp -r ./zookeeper-3.4.6 root@Worker2:/usr/local
然后分别进去Worker1和Worker2更改myid为1和2
到这个时候3台机器的Zookeeper已经配置好了
5)下一步就是让Spark支持zookeeper下HA
到spark-env.sh中配置
root@Worker1:/usr/local/spark-1.6.0-bin-hadoop2.6/conf# vi spark-env.sh
//整个集群的状态的维护和恢复都是通过zookeeper的，状态信息都是（下面的这段就是上面被注释的东西，要切单机和集群就靠这个了）
export SPARK_DAEMON_JAVA_OPTS="-Dspark.deploy.recoveryMode=ZOOKEEPER -Dspark.deploy.zookeeper.url=Master:2181,Worker1:2181,Worker2:2181 -Dspark.deploy.zookeeper.dir=/spark"
已经配置集群了，所以还要注释
#export SPARK_MASTER_IP=Master
root@Worker1:/usr/local/spark-1.6.0-bin-hadoop2.6/conf# scp spark-env.sh root@Worker1:/usr/local/spark/spark-1.6.0-bin-hadoop2.6/conf/spark-env.sh
spark-env.sh 100% 500 0.5KB/s 00:00
root@Worker1:/usr/local/spark-1.6.0-bin-hadoop2.6/conf# scp spark-env.sh root@Worker2:/usr/local/spark /spark-1.6.0-bin-hadoop2.6/conf/spark-env.sh
spark-env.sh 100% 500 0.5KB/s 00:00
到这个时候3台机器的Spark也已经配置好了，下面就是启动了
6）整体启动步骤
启动Hadoop hdfs
cd $HADOOP_HOME/sbin
./start-dfs.sh
三台Zookeeper的机器分别启动Zookeeper：
cd $ZOOKEEPER_HOME/bin
./zkServer.sh start
启动Spark
在Master启动：
cd $SPARK_HOME/sbin
./start-all.sh
./start-history-server.sh
在另外两台机器启动：
cd $SPARK_HOME/sbin
./start-mastser.sh
jps分别在三台机器上查看进程
或者看控制台
整个集群算启动好了
7）如果要实验集群效果
可以启动./spark-shell --master spark://Master:7077,Worker1:7077,Worker2:7077
然后把Master的master进程用 ./stop-master停止，过一段时间（根据机器几秒到几分钟不等）自动切换到另外的机器上
本文出自 “一枝花傲寒” 博客，请务必保留此出处http://feiweihy.blog.51cto.com/6389397/1744024
原文地址：http://feiweihy.blog.51cto.com/6389397/1744024
