storm1.0.0集群安装
1.storm集群组成
storm集群类似于Hadoop(1.x)集群,在Hadoop(1.x)集群中任务叫"MapReduce jobs"而在storm集群中任务叫"topologies","MapReduce jobs"和"topologies"的最大区别是"MapReduce jobs"执行后就结束了,但是"topologies"一直执行直到你强制关闭它。storm集群和Hadoop(1.x)集群对比如下:
Hadoop(1.x) | Storm | |
集主节点进程 | JobTracker | Nimbus |
工作节点进程 | TaskTracker | Supervisor |
应用名称 | Job | Topology |
API接口 | Mapper/Reducer | Spout/Bolt |
使用场景 | 离线数据分析处理 | 实时数据分析处理 |
storm集群中的节点分为如下三类:
- master nodes :主节点上运行的进程叫Nimbus,Nimbus主要负责分发客户端提交的代码到集群中,负责分配任务以及监控任务的执行。
- worker nodes :工作节点上运行的进程叫Supervisor,Supervisor监听分发给自己的任务代码,启动和关闭工作进程来执行Nimbus分发的任务,每个工作进程执行有很多个topology组成的任务,一个被执行的topology由许多分布于集群机器上的工作进程组成。
- zookeeper nodes: Nimbus和Supervisor节点之间所有的协调工作是通过Zookeeper集群来实现的。此外,Nimbus和Supervisor进程都是快速失败(fail-fast)和无状态(stateless)的;Storm集群所有的状态要么在Zookeeper集群中,要么存储在本地磁盘上。这意味着你可以用kill -9来杀死Nimbus和Supervisor进程,它们在重启后可以继续工作。这个设计使得Storm集群拥有不可思议的稳定性。
2.storm集群搭建
- 搭建Zookeeper集群
- 安装Storm依赖库
- 下载并解压Storm发布版本
- 修改storm.yaml配置文件
- 启动Storm各个后台进程
2.1 搭建Zookeeper集群
1) 下载并解压zookeeper3.4.6
#下载zookeeper-3.4.6.tar.gz到/opt并解压解压 cd /opt tar -zxvf zookeeper-3.4.6.tar.gz
2) 在集群中各个节点中配置/etc/hosts,内容如下:
127.0.0.1 localhost localhost.localdomain localhost4 localhost4.localdomain4 ::1 localhost localhost.localdomain localhost6 localhost6.localdomain6 192.168.202.131 master 192.168.202.132 slavery01 192.168.202.133 slavery02
3) 在集群中各个节点中创建zookeeper数据文件
sudo rm -r /home/hadoop/zookeeper cd /home/hadoop mkdir zookeeper
4) 在hostname=master机器上配置zoo.cfg,将/opt/zookeeper-3.3.4/conf目录下面的 zoo_sample.cfg修改为zoo.cfg,配置文件内容如下所示:
initLimit=10 syncLimit=5 dataDir=/home/hadoopmanage/zookeeper clientPort=2181 server.1=master:2888:3888 server.2=slavery01:2888:3888 server.3=slavery02:2888:3888
5) 远程复制分发安装文件
scp -r /opt/zookeeper-3.3.4 hadoop@slavery01:/opt/ scp -r /opt/zookeeper-3.3.4 hadoop@slavery02:/opt/
6) 在集群中各个节点设置myid必须为数字
#使用ssh是首先保证你集群中节点都做了免秘钥登录 ssh master echo "1" > /home/hadoop/zookeeper/myid ssh slavery01 echo "2" > /home/hadoop/zookeeper/myid ssh slavery02 echo "3" > /home/hadoop/zookeeper/myid
7) 启动ZooKeeper集群
cd /opt/zookeeper-3.4.6 bin/zkServer.sh start
8) 查看单机ZooKeeper是leader还是follower
cd /opt/zookeeper-3.4.6 bin/zkServer.sh status
9) 停止ZooKeeper集群
cd /opt/zookeeper-3.4.6 bin/zkServer.sh stop
10) 用客户端查看zookeeper上数据
cd /opt/zookeeper-3.4.6/ bin/zkCli.sh -server master:2181,slavery01:2181,slavery02:2181
[hadoop@master storm-1.0.0]$ cd /opt/zookeeper-3.4.6/ [hadoop@master zookeeper-3.4.6]$ bin/zkCli.sh -server master:2181,slavery01:2181,slavery02:2181 Connecting to master:2181,slavery01:2181,slavery02:2181 2016-05-02 16:39:29,880 [myid:] - INFO [main:Environment@100] - Client environment:zookeeper.version=3.4.6-1569965, built on 02/20/2014 09:09 GMT 2016-05-02 16:39:29,889 [myid:] - INFO [main:Environment@100] - Client environment:host.name=master 2016-05-02 16:39:29,889 [myid:] - INFO [main:Environment@100] - Client environment:java.version=1.7.0_65 2016-05-02 16:39:29,902 [myid:] - INFO [main:Environment@100] - Client environment:java.vendor=Oracle Corporation 2016-05-02 16:39:29,903 [myid:] - INFO [main:Environment@100] - Client environment:java.home=/opt/jdk1.7.0_65/jre 2016-05-02 16:39:29,903 [myid:] - INFO [main:Environment@100] - Client environment:java.class.path=/opt/zookeeper-3.4.6/bin/../build/classes:/opt/zookeeper-3.4.6/bin/../build/lib/*.jar:/opt/zookeeper-3.4.6/bin/../lib/slf4j-log4j12-1.6.1.jar:/opt/zookeeper-3.4.6/bin/../lib/slf4j-api-1.6.1.jar:/opt/zookeeper-3.4.6/bin/../lib/netty-3.7.0.Final.jar:/opt/zookeeper-3.4.6/bin/../lib/log4j-1.2.16.jar:/opt/zookeeper-3.4.6/bin/../lib/jline-0.9.94.jar:/opt/zookeeper-3.4.6/bin/../zookeeper-3.4.6.jar:/opt/zookeeper-3.4.6/bin/../src/java/lib/*.jar:/opt/zookeeper-3.4.6/bin/../conf:.:/opt/jdk1.7.0_65/lib/dt.jar:/opt/jdk1.7.0_65/lib/tools.jar 2016-05-02 16:39:29,903 [myid:] - INFO [main:Environment@100] - Client environment:java.library.path=/opt/hadoop-2.7.1/lib/native/:/usr/java/packages/lib/amd64:/usr/lib64:/lib64:/lib:/usr/lib 2016-05-02 16:39:29,904 [myid:] - INFO [main:Environment@100] - Client environment:java.io.tmpdir=/tmp 2016-05-02 16:39:29,904 [myid:] - INFO [main:Environment@100] - Client environment:java.compiler=2016-05-02 16:39:29,904 [myid:] - INFO [main:Environment@100] - Client environment:os.name=Linux 2016-05-02 16:39:29,904 [myid:] - INFO [main:Environment@100] - Client environment:os.arch=amd64 2016-05-02 16:39:29,905 [myid:] - INFO [main:Environment@100] - Client environment:os.version=2.6.32-358.el6.x86_64 2016-05-02 16:39:29,905 [myid:] - INFO [main:Environment@100] - Client environment:user.name=hadoop 2016-05-02 16:39:29,905 [myid:] - INFO [main:Environment@100] - Client environment:user.home=/home/hadoop 2016-05-02 16:39:29,906 [myid:] - INFO [main:Environment@100] - Client environment:user.dir=/opt/zookeeper-3.4.6 2016-05-02 16:39:29,909 [myid:] - INFO [main:ZooKeeper@438] - Initiating client connection, connectString=master:2181,slavery01:2181,slavery02:2181 sessionTimeout=30000 watcher=org.apache.zookeeper.ZooKeeperMain$MyWatcher@8afbefd Welcome to ZooKeeper! 2016-05-02 16:39:30,290 [myid:] - INFO [main-SendThread(master:2181):ClientCnxn$SendThread@975] - Opening socket connection to server master/192.168.202.131:2181. Will not attempt to authenticate using SASL (unknown error) 2016-05-02 16:39:30,350 [myid:] - INFO [main-SendThread(master:2181):ClientCnxn$SendThread@852] - Socket connection established to master/192.168.202.131:2181, initiating session JLine support is enabled 2016-05-02 16:39:31,469 [myid:] - INFO [main-SendThread(master:2181):ClientCnxn$SendThread@1235] - Session establishment complete on server master/192.168.202.131:2181, sessionid = 0x154701cef030003, negotiated timeout = 30000 WATCHER:: WatchedEvent state:SyncConnected type:None path:null [zk: master:2181,slavery01:2181,slavery02:2181(CONNECTED) 0]
查看zookeeper数据根目/录和数据目录/storm下数据:
[zk: master:2181,slavery01:2181,slavery02:2181(CONNECTED) 0] ls / [storm, hbase, zookeeper] [zk: master:2181,slavery01:2181,slavery02:2181(CONNECTED) 1] ls /storm [backpressure, workerbeats, nimbuses, supervisors, errors, logconfigs, storms, assignments, leader-lock, blobstore] [zk: master:2181,slavery01:2181,slavery02:2181(CONNECTED) 2]
2.2 安装Storm依赖库
1) jdk安装(官网要求是1.6或者以上版本,这里安装1.7)
1.卸载linux自带的jdk环境 1)首先用命令java -version 查看系统中原有的java版本 2)然后用用 rpm -qa | gcj 命令查看具体的信息 3)最后用 rpm -e --nodeps java-1.5.0-gcj-1.5.0.0-29.1.el6.x86_64卸载 2.安装jdk-7u65-linux-x64.gz 1)下载jdk-7u65-linux-x64.gz放置于/opt/java/jdk-7u65-linux-x64.gz 2)解压,输入命令tar -zxvf jdk-7u65-linux-x64.gz 3)编辑vi /etc/profile,在文件末尾追加如下内容 export JAVA_HOME=/opt/java/jdk1.7.0_65 export CLASSPATH=.:$JAVA_HOME/jre/lib/rt.jar:$JAVA_HOME/lib/dt.jar:$JAVA_HOME/lib/tools.jar export PATH=$PATH:$JAVA_HOME/bin 4)使配置生效,输入命令,source /etc/profile 5)输入命令java -version,检查JDK环境是否配置成功 2.2.2 Python安装
2) Python安装(官网要求2.6.6以上)
#安装python2.7.10 #1)在https://www.python.org/ftp/python/2.7.10/Python-2.7.10.tgz下载Python-2.7.10.tgz放置于/opt下 #2)解压tar -xzf Python-2.7.10.tgz到/opt下 cd /opt tar -xzf Python-2.7.10.tgz #3)编译安装Python cd /opt/Python-2.7.10 ./configue, make make install #4)如果不安装或者python版本过老,在后面会报错如下: #No module named argparse #5)Python版本查看 python -V
2.3 下载并解压Storm发布版本
1) 在hostname=master机器上下载Storm发行版本
cd /opt wget http://www.apache.org/dyn/closer.lua/storm/apache-storm-1.0.0/apache-storm-1.0.0.tar.gz
2) 在hostname=master机器上解压到目录/opt下
cd /opt tar -zxvf apache-storm-1.0.0.tar.gz mv apache-storm-1.0.0 storm-1.0.0
3) 在hostname=master机器上修改/opt/storm-1.0.0/conf/storm.yaml配置文件
storm.zookeeper.servers: - "master" - "slavery01" - "slavery02" nimbus.seeds: ["master"] supervisor.slots.ports: - 6700 - 6701 - 6702 - 6703 storm.local.dir: "/home/hadoopmanage/storm/localdir/"
注意:上面配置参数前的空格符、TAB符不要去掉,否则会报如下错误信息:
at org.apache.storm.shade.org.yaml.snakeyaml.scanner.ScannerImpl.stalePossibleSimpleKeys(ScannerImpl.java:460) at org.apache.storm.shade.org.yaml.snakeyaml.scanner.ScannerImpl.needMoreTokens(ScannerImpl.java:280) at org.apache.storm.shade.org.yaml.snakeyaml.scanner.ScannerImpl.checkToken(ScannerImpl.java:225) at org.apache.storm.shade.org.yaml.snakeyaml.parser.ParserImpl$ParseIndentlessSequenceEntry.produce(ParserImpl.java:532) at org.apache.storm.shade.org.yaml.snakeyaml.parser.ParserImpl.peekEvent(ParserImpl.java:158) at org.apache.storm.shade.org.yaml.snakeyaml.parser.ParserImpl.checkEvent(ParserImpl.java:143) at org.apache.storm.shade.org.yaml.snakeyaml.composer.Composer.composeSequenceNode(Composer.java:203) at org.apache.storm.shade.org.yaml.snakeyaml.composer.Composer.composeNode(Composer.java:157) at org.apache.storm.shade.org.yaml.snakeyaml.composer.Composer.composeMappingNode(Composer.java:237) at org.apache.storm.shade.org.yaml.snakeyaml.composer.Composer.composeNode(Composer.java:159) at org.apache.storm.shade.org.yaml.snakeyaml.composer.Composer.composeDocument(Composer.java:122) at org.apache.storm.shade.org.yaml.snakeyaml.composer.Composer.getSingleNode(Composer.java:105) at org.apache.storm.shade.org.yaml.snakeyaml.constructor.BaseConstructor.getSingleData(BaseConstructor.java:120) at org.apache.storm.shade.org.yaml.snakeyaml.Yaml.loadFromReader(Yaml.java:481) at org.apache.storm.shade.org.yaml.snakeyaml.Yaml.load(Yaml.java:424) at org.apache.storm.utils.Utils.findAndReadConfigFile(Utils.java:290) at org.apache.storm.utils.Utils.readStormConfig(Utils.java:391) at org.apache.storm.utils.Utils.(Utils.java:119) ... 39 more
4) 在hostname=master机器上分发安装文件到其他节点
cd /opt scp -r storm-1.0.0 hadoop@slavery01:/opt scp -r storm-1.0.0 hadoop@slavery02:/opt
5) 在各个节点上增加storm集群本地存储文件,该目录被Nimbus和Supervisor进程用于存储少量状态,如jars、confs等的本地磁盘目录,需要提前创建该目录并给以足够的访问权限。然后在storm.yaml中配置该目录
mkdir -p /home/hadoopmanage/storm/localdir/
2.4 启动Storm各个后台进程
1) 在hostname=master节点启动Nimbus进程服务,放置于后台运行
cd /opt/storm-1.0.0/ bin/storm nimbus >/dev/null 2>&1 &
2) 在各个hostname-slavery0*节点启动Supervisor
cd /opt/storm-1.0.0/ bin/storm supervisor >/dev/null 2>&1 &
3)在hostname=master节点启动UI进程服务,放置于后台运行
cd /opt/storm-1.0.0/ bin/storm ui >/dev/null 2>&1 &
启动后打开浏览器,访问http://master:8080/index.html 或者http://192.168.202.131:8080/index.html,打开的storm ui界面如下:
3.向storm集群中提交任务
3.1 启动Storm Topology
storm jar mycode.jar com.test.MyTopology arg1 arg2 arg3
其中mycode.jar是包含Topology实现代码的jar包,com.test.MyTopology的main方法是Topology的入口,arg1、arg2和arg3为org.me.MyTopology执行时需要传入的参数。
3.2 停止Storm Topology
storm kill {toponame}
其中,{toponame}为Topology提交到Storm集群时指定的Topology任务名称。