1.获取所有topic
package com.example.demo;
import java.io.IOException;
import java.util.List;
import org.apache.zookeeper.KeeperException;
import org.apache.zookeeper.WatchedEvent;
import org.apache.zookeeper.Watcher;
import org.apache.zookeeper.ZooKeeper;
public class zookeeper {
public static void main(String[] args) {
String connectString = "172.16.10.211:2181";
int sessionTimeout = 4000;
Watcher watcher = new Watcher() {
public void process(WatchedEvent event) {
}
};
try {
ZooKeeper zooKeeper = new ZooKeeper(connectString, sessionTimeout, watcher);
List<String> list = zooKeeper.getChildren("/brokers/topics", false);
int len = list.size();
for(int i = 1;i < len;i++){
System.out.println(list.get(i));
//此处动态生成消费者 //JavaKafkaConsumerHighAPI example = new JavaKafkaConsumerHighAPI(list.get(i), 1);
//new Thread(example).start();
}
} catch (IOException e) {
e.printStackTrace();
} catch (KeeperException e) {
e.printStackTrace();
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}
2.参考http://www.cnblogs.com/liuming1992/p/6432626.html生成消费者,这里进行了小小的改造
package com.example.demo;
import kafka.consumer.*;
import kafka.javaapi.consumer.ConsumerConnector;
import kafka.message.MessageAndMetadata;
import kafka.serializer.StringDecoder;
import kafka.utils.VerifiableProperties;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Properties;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;
/**
* 自定义简单Kafka消费者, 使用高级API
* Created by gerry on 12/21.
*/
public class JavaKafkaConsumerHighAPI implements Runnable {
/**
* Kafka数据消费对象
*/
private ConsumerConnector consumer;
/**
* Kafka Topic名称
*/
private String topic;
/**
* 线程数量,一般就是Topic的分区数量
*/
private int numThreads;
/**
* 线程池
*/
private ExecutorService executorPool;
/**
* 构造函数
*
* @param topic Kafka消息Topic主题
* @param numThreads 处理数据的线程数/可以理解为Topic的分区数
* @param zookeeper Kafka的Zookeeper连接字符串
* @param groupId 该消费者所属group ID的值
*/
public JavaKafkaConsumerHighAPI(String topic, int numThreads) {
// 1. 创建Kafka连接器
this.consumer = Consumer.createJavaConsumerConnector(createConsumerConfig("172.16.10.211:2181", "test-consumer-group"));
// 2. 数据赋值
this.topic = topic;
this.numThreads = numThreads;
}
@Override
public void run() {
// 1. 指定Topic
Map<String, Integer> topicCountMap = new HashMap<String, Integer>();
topicCountMap.put(this.topic, this.numThreads);
// 2. 指定数据的解码器
StringDecoder keyDecoder = new StringDecoder(new VerifiableProperties());
StringDecoder valueDecoder = new StringDecoder(new VerifiableProperties());
// 3. 获取连接数据的迭代器对象集合
/**
* Key: Topic主题
* Value: 对应Topic的数据流读取器,大小是topicCountMap中指定的topic大小
*/
Map<String, List<KafkaStream<String, String>>> consumerMap = this.consumer.createMessageStreams(topicCountMap, keyDecoder, valueDecoder);
// 4. 从返回结果中获取对应topic的数据流处理器
List<KafkaStream<String, String>> streams = consumerMap.get(this.topic);
// 5. 创建线程池
this.executorPool = Executors.newFixedThreadPool(this.numThreads);
// 6. 构建数据输出对象
int threadNumber = 0;
for (final KafkaStream<String, String> stream : streams) {
this.executorPool.submit(new ConsumerKafkaStreamProcesser(stream, threadNumber,topic));
threadNumber++;
}
}
public void shutdown() {
// 1. 关闭和Kafka的连接,这样会导致stream.hashNext返回false
if (this.consumer != null) {
this.consumer.shutdown();
}
// 2. 关闭线程池,会等待线程的执行完成
if (this.executorPool != null) {
// 2.1 关闭线程池
this.executorPool.shutdown();
// 2.2. 等待关闭完成, 等待五秒
try {
if (!this.executorPool.awaitTermination(5, TimeUnit.SECONDS)) {
System.out.println("Timed out waiting for consumer threads to shut down, exiting uncleanly!!");
}
} catch (InterruptedException e) {
System.out.println("Interrupted during shutdown, exiting uncleanly!!");
}
}
}
/**
* 根据传入的zk的连接信息和groupID的值创建对应的ConsumerConfig对象
*
* @param zookeeper zk的连接信息,类似于:<br/>
* hadoop-senior01.ibeifeng.com:2181,hadoop-senior02.ibeifeng.com:2181/kafka
* @param groupId 该kafka consumer所属的group id的值, group id值一样的kafka consumer会进行负载均衡
* @return Kafka连接信息
*/
private ConsumerConfig createConsumerConfig(String zookeeper, String groupId) {
// 1. 构建属性对象
Properties prop = new Properties();
// 2. 添加相关属性
prop.put("group.id", groupId); // 指定分组id
prop.put("zookeeper.connect", zookeeper); // 指定zk的连接url
prop.put("zookeeper.session.timeout.ms", "400"); //
prop.put("zookeeper.sync.time.ms", "200");
prop.put("auto.commit.interval.ms", "1000");
// 3. 构建ConsumerConfig对象
return new ConsumerConfig(prop);
}
/**
* Kafka消费者数据处理线程
*/
public static class ConsumerKafkaStreamProcesser implements Runnable {
// Kafka数据流
private KafkaStream<String, String> stream;
// 线程ID编号
private int threadNumber;
private String topic;
public ConsumerKafkaStreamProcesser(KafkaStream<String, String> stream, int threadNumber,String topic) {
this.stream = stream;
this.threadNumber = threadNumber;
this.topic = topic;
}
@Override
public void run() {
// 1. 获取数据迭代器
ConsumerIterator<String, String> iter = this.stream.iterator();
// 2. 迭代输出数据
while (iter.hasNext()) {
// 2.1 获取数据值
MessageAndMetadata value = iter.next();
// 2.2 输出
System.out.println(this.threadNumber + "____" + value.offset() +"_____"+ topic + "____" + value.message());
}
// 3. 表示当前线程执行完成
System.out.println("Shutdown Thread:" + this.threadNumber);
}
}
}
3.pom
<dependency>
<groupId>org.apache.kafka</groupId>
<artifactId>kafka_2.11</artifactId>
<version>0.8.2.1</version>
</dependency>