Spark集群基于Zookeeper的HA搭建部署

1.环境介绍
（1）操作系统RHEL6.2-64
（2）两个节点：spark1(192.168.232.147),spark2(192.168.232.152)
（3）两个节点上都装好了Hadoop 2.2集群
2.安装Zookeeper
（1）下载Zookeeper：http://apache.claz.org/zookeeper ... keeper-3.4.5.tar.gz
（2）解压到/root/install/目录下
（3）创建两个目录，一个是数据目录，一个日志目录
 
（4）配置：进到conf目录下，把zoo_sample.cfg修改成zoo.cfg（这一步是必须的，否则zookeeper不认识zoo_sample.cfg），并添加如下内容

1. dataDir=/root/install/zookeeper-3.4.5/data
   
2. dataLogDir=/root/install/zookeeper-3.4.5/logs
   
3. server.1=spark1:2888:3888
   
4. server.2=spark2:2888:3888

复制代码

（5）在/root/install/zookeeper-3.4.5/data目录下创建myid文件，并在里面写1

1. cd /root/install/zookeeper-3.4.5/data
   
2. echo 1>myid

复制代码

（6）把/root/install/zookeeper-3.4.5整个目录复制到其他节点

1. scp -r /root/install/zookeeper-3.4.5 root@spark2:/root/install/

复制代码

（7）登录到spark2节点，修改myid文件里的值，将其修改为2

1. cd /root/install/zookeeper-3.4.5/data
   
2. echo 2>myid

复制代码

（8）在spark1,spark2两个节点上分别启动zookeeper

1. cd /root/install/zookeeper-3.4.5
   
2. bin/zkServer.sh start

复制代码

（9）查看进程进否成在

1. [root@spark2 zookeeper-3.4.5]# bin/zkServer.sh start
   
2. JMX enabled by default
   
3. Using config: /root/install/zookeeper-3.4.5/bin/../conf/zoo.cfg
   
4. Starting zookeeper ... STARTED
   
5. [root@spark2 zookeeper-3.4.5]# jps
   
6. 2490 Jps
   
7. 2479 QuorumPeerMain

复制代码

3.配置Spark的HA
（1）进到spark的配置目录，在spark-env.sh修改如下

1. export SPARK_DAEMON_JAVA_OPTS="-Dspark.deploy.recoveryMode=ZOOKEEPER -Dspark.deploy.zookeeper.url=spark1:2181,spark2:2181 -Dspark.deploy.zookeeper.dir=/spark"
   
2. export JAVA_HOME=/root/install/jdk1.7.0_21
   
3. #export SPARK_MASTER_IP=spark1
   
4. #export SPARK_MASTER_PORT=7077
   
5. export SPARK_WORKER_CORES=1
   
6. export SPARK_WORKER_INSTANCES=1
   
7. export SPARK_WORKER_MEMORY=1g

复制代码

（2）把这个配置文件分发到各个节点上去

1. scp spark-env.sh root@spark2:/root/install/spark-1.0/conf/

复制代码

（3）启动spark集群

1. [root@spark1 spark-1.0]# sbin/start-all.sh 
   
2. starting org.apache.spark.deploy.master.Master, logging to /root/install/spark-1.0/sbin/../logs/spark-root-org.apache.spark.deploy.master.Master-1-spark1.out
   
3. spark1: starting org.apache.spark.deploy.worker.Worker, logging to /root/install/spark-1.0/sbin/../logs/spark-root-org.apache.spark.deploy.worker.Worker-1-spark1.out
   
4. spark2: starting org.apache.spark.deploy.worker.Worker, logging to /root/install/spark-1.0/sbin/../logs/spark-root-org.apache.spark.deploy.worker.Worker-1-spark2.out

复制代码

（4）进到spark2(192.168.232.152)节点，把start-master.sh 启动，当spark1(192.168.232.147)挂掉时，spark2顶替当master

1. [root@spark2 spark-1.0]# sbin/start-master.sh 
   
2. starting org.apache.spark.deploy.master.Master, logging to /root/install/spark-1.0/sbin/../logs/spark-root-org.apache.spark.deploy.master.Master-1-spark2.out
   

复制代码

（5）查看spark1和spark2上运行的哪些进程

1. [root@spark1 spark-1.0]# jps
   
2. 5797 Worker
   
3. 5676 Master
   
4. 6287 Jps
   
5. 2602 QuorumPeerMain
   
6. [root@spark2 spark-1.0]# jps
   
7. 2479 QuorumPeerMain
   
8. 5750 Jps
   
9. 5534 Worker
   
10. 5635 Master
    

复制代码

4.测试HA是否生效
（1）先查看一下两个节点的运行情况，现在spark1运行了master,spark2是待命状态
 
（2）在spark1上把master服务停掉

1. [root@spark1 spark-1.0]# sbin/stop-master.sh 
   
2. stopping org.apache.spark.deploy.master.Master
   
3. [root@spark1 spark-1.0]# jps
   
4. 5797 Worker
   
5. 6373 Jps
   
6. 2602 QuorumPeerMain

复制代码

（3）用浏览器访问master的8080端口，看是否还活着。以下可以看出，master已经挂掉
 
（4）再用浏览器访问查看spark2的状态，从下图看出，spark2已经被切换当master了