kafka_2.9.2-0.8.1.1分布式集群搭建代码开发实例

准备3台虚拟机, 系统是RHEL64服务版. 1) 每台机器配置如下:
$ cat /etc/hosts
    # zookeeper hostnames:       192.168.8.182       zk1       192.168.8.183       zk2       192.168.8.184       zk3  
2) 每台机器上安装jdk, zookeeper, kafka, 配置如下:
$ vi /etc/profile            # jdk, zookeeper, kafka       export KAFKA_HOME=/usr/local/lib/kafka/kafka_2.9.2-0.8.11       export ZK_HOME=/usr/local/lib/zookeeper/zookeeper-3.4.6       export CLASSPATH=.:$JAVA_HOME/lib/tools.jar:$JAVA_HOME/lib/dt.jar       export PATH=$JAVA_HOME/bin:$JAVA_HOME/jre/bin:$KAFKA_HOME/bin:$ZK_HOME/bin:$PATH  
3) 每台机器上运行:
$ source /etc/profile
$ mkdir -p /var/lib/zookeeper
$ cd $ZK_HOME/conf
$ cp zoo_sample.cfg zoo.cfg
$ vi zoo.cfg            dataDir=/var/lib/zookeeper              # the port at which the clients will connect       clientPort=2181              # zookeeper cluster       server.1=zk1:2888:3888       server.2=zk2:2888:3888       server.3=zk3:2888:3888  
4) 每台机器上生成myid:
zk1:
$ echo "1" > /var/lib/zookeeper/myid
zk2:
$ echo "2" > /var/lib/zookeeper/myid
zk3:
$ echo "3" > /var/lib/zookeeper/myid 5) 每台机器上运行setup关闭防火墙
Firewall:
[   ] enabled 6) 每台机器上启动zookeeper:
$ zkServer.sh start
查看状态:
$ zkServer.sh status
1)下载KAFKA
    $ wget http://apache.fayea.com/apache-mirror/kafka/0.8.1.1/kafka_2.9.2-0.8.1.1.tgz
安装和配置参考上一篇文章:
http://blog.csdn.net/ubuntu64fan/article/details/26678877
2)配置$KAFKA_HOME/config/server.properties
我们安装3个broker，分别在3个vm上：zk1，zk2，zk3:
zk1:
$ vi /etc/sysconfig/network
    NETWORKING=yes       HOSTNAME=zk1  
$ vi $KAFKA_HOME/config/server.properties
    broker.id=0       port=9092       host.name=zk1       advertised.host.name=zk1       ...       num.partitions=2       ...       zookeeper.contact=zk1:2181,zk2:2181,zk3:2181  
zk2:
$ vi /etc/sysconfig/network
    NETWORKING=yes       HOSTNAME=zk2  
$ vi $KAFKA_HOME/config/server.properties
    broker.id=1       port=9092       host.name=zk2       advertised.host.name=zk2       ...       num.partitions=2       ...       zookeeper.contact=zk1:2181,zk2:2181,zk3:2181  
zk3:
$ vi /etc/sysconfig/network
    NETWORKING=yes       HOSTNAME=zk3  
$ vi $KAFKA_HOME/config/server.properties
    broker.id=2       port=9092       host.name=zk3       advertised.host.name=zk3       ...       num.partitions=2       ...       zookeeper.contact=zk1:2181,zk2:2181,zk3:2181  
3)启动zookeeper服务, 在zk1,zk2,zk3上分别运行：
$ zkServer.sh start 4)启动kafka服务, 在zk1,zk2,zk3上分别运行：
$ kafka-server-start.sh $KAFKA_HOME/config/server.properties 5) 新建一个TOPIC(replication-factor=num of brokers)
$ kafka-topics.sh --create --topic test --replication-factor 3 --partitions 2 --zookeeper zk1:2181 6)假设我们在zk2上，开一个终端，发送消息至kafka（zk2模拟producer）
$ kafka-console-producer.sh --broker-list zk1:9092 --sync --topic test
在发送消息的终端输入：Hello Kafka
7)假设我们在zk3上，开一个终端，显示消息的消费（zk3模拟consumer）
$ kafka-console-consumer.sh --zookeeper zk1:2181 --topic test --from-beginning 在消费消息的终端显示：Hello Kafka
项目准备开发
项目基于maven构建，不得不说kafka java客户端实在是太糟糕了；构建环境会遇到很多麻烦。建议参考如下pom.xml;其中各个依赖包必须版本协调一致。如果kafka client的版
本和kafka server的版本不一致,将会有很多异常,比如"broker id not exists"等;因为kafka从0.7升级到0.8之后(正名为2.8.0),client与server通讯的protocol已经改变.
  

Xml代码  收藏代码
  1. <dependencies>   
  2.        <dependency>   
  3.            <groupId>log4j</groupId>   
  4.            <artifactId>log4j</artifactId>   
  5.            <version>1.2.14</version>   
  6.        </dependency>   
  7.        <dependency>   
  8.            <groupId>org.apache.kafka</groupId>   
  9.            <artifactId>kafka_2.8.2</artifactId>   
  10.            <version>0.8.0</version>   
  11.            <exclusions>   
  12.                <exclusion>   
  13.                    <groupId>log4j</groupId>   
  14.                    <artifactId>log4j</artifactId>   
  15.                </exclusion>   
  16.            </exclusions>   
  17.        </dependency>   
  18.        <dependency>   
  19.            <groupId>org.scala-lang</groupId>   
  20.            <artifactId>scala-library</artifactId>   
  21.            <version>2.8.2</version>   
  22.        </dependency>   
  23.        <dependency>   
  24.            <groupId>com.yammer.metrics</groupId>   
  25.            <artifactId>metrics-core</artifactId>   
  26.            <version>2.2.0</version>   
  27.        </dependency>   
  28.        <dependency>   
  29.            <groupId>com.101tec</groupId>   
  30.            <artifactId>zkclient</artifactId>   
  31.            <version>0.3</version>   
  32.        </dependency>   
  33.    </dependencies>    

 
Producer端代码
    1) producer.properties文件：此文件放在/resources目录下
  

Xml代码  收藏代码
  1. #partitioner.class=   
  2.    ##broker列表可以为kafka server的子集,因为producer需要从broker中获取metadata   
  3.    ##尽管每个broker都可以提供metadata,此处还是建议,将所有broker都列举出来   
  4.    ##此值,我们可以在spring中注入过来   
  5.    ##metadata.broker.list=127.0.0.1:9092,127.0.0.1:9093   
  6.    ##,127.0.0.1:9093   
  7.    ##同步,建议为async   
  8.    producer.type=sync   
  9.    compression.codec=0   
  10.    serializer.class=kafka.serializer.StringEncoder   
  11.    ##在producer.type=async时有效   
  12.    #batch.num.messages=100    

 
    2) KafkaProducerClient.java代码样例
  

Java代码  收藏代码
  1. import java.util.ArrayList;   
  2.    import java.util.Collection;   
  3.    import java.util.List;   
  4.    import java.util.Properties;   
  5.       
  6.    import kafka.javaapi.producer.Producer;   
  7.    import kafka.producer.KeyedMessage;   
  8.    import kafka.producer.ProducerConfig;   
  9.       
  10.    public class KafkaProducerClient {   
  11.       
  12.        private Producer<String, String> inner;   
  13.           
  14.        private String brokerList;//for metadata discovery,spring setter   
  15.        private String location = "kafka-producer.properties";//spring setter   
  16.           
  17.        private String defaultTopic;//spring setter   
  18.       
  19.        public void setBrokerList(String brokerList) {   
  20.            this.brokerList = brokerList;   
  21.        }   
  22.       
  23.        public void setLocation(String location) {   
  24.            this.location = location;   
  25.        }   
  26.       
  27.        public void setDefaultTopic(String defaultTopic) {   
  28.            this.defaultTopic = defaultTopic;   
  29.        }   
  30.       
  31.        public KafkaProducerClient(){}   
  32.           
  33.        public void init() throws Exception {   
  34.            Properties properties = new Properties();   
  35.            properties.load(Thread.currentThread().getContextClassLoader().getResourceAsStream(location));   
  36.               
  37.               
  38.            if(brokerList != null) {   
  39.                properties.put("metadata.broker.list", brokerList);   
  40.            }   
  41.       
  42.            ProducerConfig config = new ProducerConfig(properties);   
  43.            inner = new Producer<String, String>(config);   
  44.        }   
  45.       
  46.        public void send(String message){   
  47.            send(defaultTopic,message);   
  48.        }   
  49.           
  50.        public void send(Collection<String> messages){   
  51.            send(defaultTopic,messages);   
  52.        }   
  53.           
  54.        public void send(String topicName, String message) {   
  55.            if (topicName == null || message == null) {   
  56.                return;   
  57.            }   
  58.            KeyedMessage<String, String> km = new KeyedMessage<String, String>(topicName,message);   
  59.            inner.send(km);   
  60.        }   
  61.       
  62.        public void send(String topicName, Collection<String> messages) {   
  63.            if (topicName == null || messages == null) {   
  64.                return;   
  65.            }   
  66.            if (messages.isEmpty()) {   
  67.                return;   
  68.            }   
  69.            List<KeyedMessage<String, String>> kms = new ArrayList<KeyedMessage<String, String>>();   
  70.            int i= 0;   
  71.            for (String entry : messages) {   
  72.                KeyedMessage<String, String> km = new KeyedMessage<String, String>(topicName,entry);   
  73.                kms.add(km);   
  74.                i++;   
  75.                if(i % 20 == 0){   
  76.                    inner.send(kms);   
  77.                    kms.clear();   
  78.                }   
  79.            }   
  80.               
  81.            if(!kms.isEmpty()){   
  82.                inner.send(kms);   
  83.            }   
  84.        }   
  85.       
  86.        public void close() {   
  87.            inner.close();   
  88.        }   
  89.       
  90.        /** 
  91.         * @param args 
  92.         */   
  93.        public static void main(String[] args) {   
  94.            KafkaProducerClient producer = null;   
  95.            try {   
  96.                producer = new KafkaProducerClient();   
  97.                //producer.setBrokerList("");   
  98.                int i = 0;   
  99.                while (true) {   
  100.                    producer.send("test-topic", "this is a sample" + i);   
  101.                    i++;   
  102.                    Thread.sleep(2000);   
  103.                }   
  104.            } catch (Exception e) {   
  105.                e.printStackTrace();   
  106.            } finally {   
  107.                if (producer != null) {   
  108.                    producer.close();   
  109.                }   
  110.            }   
  111.       
  112.        }   
  113.       
  114.    }   

  Consumer端
     1) consumer.properties:文件位于/resources目录下

Xml代码  收藏代码
  1. ## 此值可以配置,也可以通过spring注入   
  2.    ##zookeeper.connect=127.0.0.1:2181,127.0.0.1:2182,127.0.0.1:2183   
  3.    ##,127.0.0.1:2182,127.0.0.1:2183   
  4.    # timeout in ms for connecting to zookeeper   
  5.    zookeeper.connectiontimeout.ms=1000000   
  6.    #consumer group id   
  7.    group.id=test-group   
  8.    #consumer timeout   
  9.    #consumer.timeout.ms=5000   
  10.    auto.commit.enable=true   
  11.    auto.commit.interval.ms=60000    

 
    2) KafkaConsumerClient.java代码样例
  

Java代码  收藏代码
  1. package com.test.kafka;   
  2.    import java.nio.ByteBuffer;   
  3.    import java.nio.CharBuffer;   
  4.    import java.nio.charset.Charset;   
  5.    import java.util.HashMap;   
  6.    import java.util.List;   
  7.    import java.util.Map;   
  8.    import java.util.Properties;   
  9.    import java.util.concurrent.ExecutorService;   
  10.    import java.util.concurrent.Executors;   
  11.       
  12.    import kafka.consumer.Consumer;   
  13.    import kafka.consumer.ConsumerConfig;   
  14.    import kafka.consumer.ConsumerIterator;   
  15.    import kafka.consumer.KafkaStream;   
  16.    import kafka.javaapi.consumer.ConsumerConnector;   
  17.    import kafka.message.Message;   
  18.    import kafka.message.MessageAndMetadata;   
  19.       
  20.    public class KafkaConsumerClient {   
  21.       
  22.        private String groupid; //can be setting by spring   
  23.        private String zkConnect;//can be setting by spring   
  24.        private String location = "kafka-consumer.properties";//配置文件位置   
  25.        private String topic;   
  26.        private int partitionsNum = 1;   
  27.        private MessageExecutor executor; //message listener   
  28.        private ExecutorService threadPool;   
  29.           
  30.        private ConsumerConnector connector;   
  31.           
  32.        private Charset charset = Charset.forName("utf8");   
  33.       
  34.        public void setGroupid(String groupid) {   
  35.            this.groupid = groupid;   
  36.        }   
  37.       
  38.        public void setZkConnect(String zkConnect) {   
  39.            this.zkConnect = zkConnect;   
  40.        }   
  41.       
  42.        public void setLocation(String location) {   
  43.            this.location = location;   
  44.        }   
  45.       
  46.        public void setTopic(String topic) {   
  47.            this.topic = topic;   
  48.        }   
  49.       
  50.        public void setPartitionsNum(int partitionsNum) {   
  51.            this.partitionsNum = partitionsNum;   
  52.        }   
  53.       
  54.        public void setExecutor(MessageExecutor executor) {   
  55.            this.executor = executor;   
  56.        }   
  57.       
  58.        public KafkaConsumerClient() {}   
  59.       
  60.        //init consumer,and start connection and listener   
  61.        public void init() throws Exception {   
  62.            if(executor == null){   
  63.                throw new RuntimeException("KafkaConsumer,exectuor cant be null!");   
  64.            }   
  65.            Properties properties = new Properties();   
  66.            properties.load(Thread.currentThread().getContextClassLoader().getResourceAsStream(location));   
  67.               
  68.            if(groupid != null){   
  69.                properties.put("groupid", groupid);   
  70.            }   
  71.            if(zkConnect != null){   
  72.                properties.put("zookeeper.connect", zkConnect);   
  73.            }   
  74.            ConsumerConfig config = new ConsumerConfig(properties);   
  75.       
  76.            connector = Consumer.createJavaConsumerConnector(config);   
  77.            Map<String, Integer> topics = new HashMap<String, Integer>();   
  78.            topics.put(topic, partitionsNum);   
  79.            Map<String, List<KafkaStream<byte[], byte[]>>> streams = connector.createMessageStreams(topics);   
  80.            List<KafkaStream<byte[], byte[]>> partitions = streams.get(topic);   
  81.            threadPool = Executors.newFixedThreadPool(partitionsNum * 2);   
  82.               
  83.            //start   
  84.            for (KafkaStream<byte[], byte[]> partition : partitions) {   
  85.                threadPool.execute(new MessageRunner(partition));   
  86.            }   
  87.        }   
  88.       
  89.        public void close() {   
  90.            try {   
  91.                threadPool.shutdownNow();   
  92.            } catch (Exception e) {   
  93.                //   
  94.            } finally {   
  95.                connector.shutdown();   
  96.            }   
  97.       
  98.        }   
  99.       
  100.        class MessageRunner implements Runnable {   
  101.            private KafkaStream<byte[], byte[]> partition;   
  102.       
  103.            MessageRunner(KafkaStream<byte[], byte[]> partition) {   
  104.                this.partition = partition;   
  105.            }   
  106.       
  107.            public void run() {   
  108.                ConsumerIterator<byte[], byte[]> it = partition.iterator();   
  109.                while (it.hasNext()) {   
  110.                    // connector.commitOffsets();手动提交offset,当autocommit.enable=false时使用   
  111.                    MessageAndMetadata<byte[], byte[]> item = it.next();   
  112.                    try{   
  113.                        executor.execute(new String(item.message(),charset));// UTF-8,注意异常   
  114.                    }catch(Exception e){   
  115.                        //   
  116.                    }   
  117.                }   
  118.            }   
  119.               
  120.            public String getContent(Message message){   
  121.                ByteBuffer buffer = message.payload();   
  122.                if (buffer.remaining() == 0) {   
  123.                    return null;   
  124.                }   
  125.                CharBuffer charBuffer = charset.decode(buffer);   
  126.                return charBuffer.toString();   
  127.            }   
  128.        }   
  129.       
  130.        public static interface MessageExecutor {   
  131.       
  132.            public void execute(String message);   
  133.        }   
  134.       
  135.        /** 
  136.         * @param args 
  137.         */   
  138.        public static void main(String[] args) {   
  139.            KafkaConsumerClient consumer = null;   
  140.            try {   
  141.                MessageExecutor executor = new MessageExecutor() {   
  142.       
  143.                    public void execute(String message) {   
  144.                        System.out.println(message);   
  145.                    }   
  146.                };   
  147.                consumer = new KafkaConsumerClient();   
  148.                   
  149.                consumer.setTopic("test-topic");   
  150.                consumer.setPartitionsNum(2);   
  151.                consumer.setExecutor(executor);   
  152.                consumer.init();   
  153.            } catch (Exception e) {   
  154.                e.printStackTrace();   
  155.            } finally {   
  156.                 if(consumer != null){   
  157.                     consumer.close();   
  158.                 }   
  159.            }   
  160.       
  161.        }   
  162.       
  163.    }    

 
    需要提醒的是,上述LogConsumer类中,没有太多的关注异常情况,必须在MessageExecutor.execute()方法中抛出异常时的情况.
    在测试时，建议优先启动consumer，然后再启动producer，这样可以实时的观测到最新的消息。

posted @ 2014-11-25 17:16  TonyChai  阅读(665)  评论(0)  编辑  收藏  举报