一、特点：
不用关心offset, 会自动的读zookeeper中该Consumer group的last offset
?
二、注意事项
1. 如果consumer比partition多，是浪费，因为kafka的设计是在一个partition上是不允许并发的，
? ?所以consumer数不要大于partition数?
2. 如果consumer比partition少，一个consumer会对应于多个partitions，
? ? 这里主要合理分配consumer数和partition数，否则会导致partition里面的数据被取的不均匀?
? ? 最好partiton数目是consumer数目的整数倍，所以partition数目很重要，
? ? 比如取24，就很容易设定consumer数目?
3. 如果consumer从多个partition读到数据，不保证数据间的顺序性，
? ? kafka只保证在一个partition上数据是有序的，但多个partition，根据你读的顺序会有不同?
4. 增减consumer，broker，partition会导致rebalance，
? ? ?所以rebalance后consumer对应的partition会发生变化?
5. High-level接口中获取不到数据的时候是会block的?
?
三、代码如下：
package kafkatest.kakfademo;
?
import java.io.UnsupportedEncodingException;
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 kafka.consumer.Consumer;
import kafka.consumer.ConsumerConfig;
import kafka.consumer.ConsumerIterator;
import kafka.consumer.KafkaStream;
import kafka.javaapi.consumer.ConsumerConnector;
?
public class ConsumerDemo1 {
public static void main(String[] args) {
ConsumerDemo1 demo = new ConsumerDemo1();
demo.test();
}
?
@SuppressWarnings("rawtypes")
public void test() {
String topicName = "test";
int numThreads = 1;
Properties properties = new Properties();
properties.put("zookeeper.connect", "hadoop0:2181");// 声明zk
properties.put("group.id", "group--demo");// 必须要使用别的组名称，
// 如果生产者和消费者都在同一组，则不能访问同一组内的topic数据
?
ConsumerConnector consumer = Consumer
.createJavaConsumerConnector(new ConsumerConfig(properties));
Map<String, Integer> topicCountMap = new HashMap<String, Integer>();
topicCountMap.put(topicName, numThreads); // 一次从主题中获取一个数据
Map<String, List<KafkaStream<byte[], byte[]>>> messageStreams = consumer
.createMessageStreams(topicCountMap);
// 获取每次接收到的这个数据
List<KafkaStream<byte[], byte[]>> streams = messageStreams
.get(topicName);
?
// now launch all the threads
ExecutorService executor = Executors.newFixedThreadPool(numThreads);
?
// now create an object to consume the messages
//
int threadNumber = 0;
for (final KafkaStream stream : streams) {
executor.execute(new ConsumerMsgTask(stream, threadNumber));
threadNumber++;
}
}
?
class ConsumerMsgTask implements Runnable {
?
private KafkaStream m_stream;
private int m_threadNumber;
?
public ConsumerMsgTask(KafkaStream stream, int threadNumber) {
m_threadNumber = threadNumber;
m_stream = stream;
}
?
public void run() {
ConsumerIterator<byte[], byte[]> it = m_stream.iterator();
long offset = 0;
try {
while (it.hasNext())
offset = it.next().offset();
byte[] bytes = it.next().message();
?
String msg = new String(bytes, "UTF-8");
?
System.out.print("offset: " + offset + ",msg:" + msg);
System.out.println("Shutting down Thread: " + m_threadNumber);
} catch (UnsupportedEncodingException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
}
}
?
}
?
?
四、实验验证
?
?
?
?
?
【大数据Kafka之 high-level--Consumer 】
原文：http://gaojingsong.iteye.com/blog/2302784