java实现Kafka的消费者示例

使用java实现Kafka的消费者

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
package com.lisg.kafkatest;
 
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;
 
import kafka.consumer.Consumer;
import kafka.consumer.ConsumerConfig;
import kafka.consumer.ConsumerIterator;
import kafka.consumer.KafkaStream;
import kafka.javaapi.consumer.ConsumerConnector;
 
/**
 * java实现Kafka消费者的示例
 * @author lisg
 *
 */
public class KafkaConsumer {
    private static final String TOPIC = "test";
    private static final int THREAD_AMOUNT = 1;
 
    public static void main(String[] args) {
         
        Properties props = new Properties();
        props.put("zookeeper.connect", "vm1:2181");
        props.put("group.id", "group1");
        props.put("zookeeper.session.timeout.ms", "400");
        props.put("zookeeper.sync.time.ms", "200");
        props.put("auto.commit.interval.ms", "1000");;
         
        Map<String, Integer> topicCountMap = new HashMap<String, Integer>();
        //每个topic使用多少个kafkastream读取, 多个consumer
        topicCountMap.put(TOPIC, THREAD_AMOUNT);
        //可以读取多个topic
//      topicCountMap.put(TOPIC2, 1);
        ConsumerConnector consumer = Consumer.createJavaConsumerConnector(new ConsumerConfig(props));
        Map<String, List<KafkaStream<byte[], byte[]>>> msgStreams = consumer.createMessageStreams(topicCountMap );
        List<KafkaStream<byte[], byte[]>> msgStreamList = msgStreams.get(TOPIC);
         
        //使用ExecutorService来调度线程
        ExecutorService executor = Executors.newFixedThreadPool(THREAD_AMOUNT);
        for (int i = 0; i < msgStreamList.size(); i++) {
            KafkaStream<byte[], byte[]> kafkaStream = msgStreamList.get(i);
            executor.submit(new HanldMessageThread(kafkaStream, i));
        }
         
         
        //关闭consumer
        try {
            Thread.sleep(20000);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        if (consumer != null) {
            consumer.shutdown();
        }
        if (executor != null) {
            executor.shutdown();
        }
        try {
            if (!executor.awaitTermination(5000, TimeUnit.MILLISECONDS)) {
                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");
        }
    }
 
}
 
/**
 * 具体处理message的线程
 * @author Administrator
 *
 */
class HanldMessageThread implements Runnable {
 
    private KafkaStream<byte[], byte[]> kafkaStream = null;
    private int num = 0;
     
    public HanldMessageThread(KafkaStream<byte[], byte[]> kafkaStream, int num) {
        super();
        this.kafkaStream = kafkaStream;
        this.num = num;
    }
 
    public void run() {
        ConsumerIterator<byte[], byte[]> iterator = kafkaStream.iterator();
        while(iterator.hasNext()) {
            String message = new String(iterator.next().message());
            System.out.println("Thread no: " + num + ", message: " + message);
        }
    }
     
}
1
props.put("auto.commit.interval.ms", "1000");

表示的是:consumer间隔多长时间在zookeeper上更新一次offset

说明:

为什么使用High Level Consumer?

有些场景下,从Kafka中读取消息的逻辑不处理消息的offset,仅仅是获取消息数据。High Level Consumer就提供了这种功能。

首先要知道的是,High Level Consumer在ZooKeeper上保存最新的offset(从指定的分区中读取)。这个offset基于consumer group名存储。

Consumer group名在Kafka集群上是全局性的,在启动新的consumer group的时候要小心集群上没有关闭的consumer。当一个consumer线程启动了,Kafka会将它加入到相同的topic下的相同consumer group里,并且触发重新分配。在重新分配时,Kafka将partition分配给consumer,有可能会移动一个partition给另一个consumer。如果老的、新的处理逻辑同时存在,有可能一些消息传递到了老的consumer上。

设计High Level Consumer

使用High LevelConsumer首先要知道的是,它应该是多线程的。消费者线程的数量跟tipic的partition数量有关,它们之间有一些特定的规则:

  • 如果线程数量大于主题的分区数量,一些线程将得不到任何消息

  • 如果分区数大于线程数,一些线程将得到多个分区的消息

  • 如果一个线程处理多个分区的消息,它接收到消息的顺序是不能保证的。比如,先从分区10获取了5条消息,从分区11获取了6条消息,然后从分区10获取了5条,紧接着又从分区10获取了5条,虽然分区11还有消息。

  • 添加更多了同consumer group的consumer将触发Kafka重新分配,某个分区本来分配给a线程的,从新分配后,有可能分配给了b线程。

关闭消费组和错误处理

Kafka不会再每次读取消息后马上更新zookeeper上的offset,而是等待一段时间。由于这种延迟,有可能消费者读取了一条消息,但没有更新offset。所以,当客户端关闭或崩溃后,从新启动时有些消息重复读取了。另外,broker宕机或其他原因导致更换了partition的leader,也会导致消息重复读取。

为了避免这种问题,你应该提供一个平滑的关闭方式,而不是使用kill -9

上面的java代码中提供一种关闭的方式:

1
2
3
4
5
6
7
8
9
10
11
12
13
if (consumer != null) {
    consumer.shutdown();
}
if (executor != null) {
    executor.shutdown();
}
try {
    if (!executor.awaitTermination(5000, TimeUnit.MILLISECONDS)) {
        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");
}

在shutdown之后,等待了5秒钟,给consumer线程时间来处理完kafka stream里保留的消息。

参考资料:https://cwiki.apache.org/confluence/display/KAFKA/Consumer+Group+Example











附件列表

    原文地址:https://www.cnblogs.com/lishouguang/p/4560561.html