2014年11月19日

by smallnest

Kafka和Spring集成实践

目录 [−]

安装Zookeeper
安装Kafka
创建一个Spring项目

本文以单机的环境演示如何将Kafka和Spring集成。
单机的环境最容易搭建，并且只需在自己的PC上运行即可，不需要很多的硬件环境，便于学习。况且，本文的目的不是搭建ZooKeeper的集群环境，而是重点介绍Kafka和Spring的应用。
具体的软件环境如下：

OS: CentOS 6.4
Zookepper： zookeeper-3.4.6
Kafka： kafka_2.9.1-0.8.2-beta
Java： JDK 1.7.0_45-b18
Spring:4.0.6

本例子在我的这个环境中运行正常，全部代码可以到 github 下载。

本文所有的操作系统用户都是root。实际产品中可能安全标准需要特定的用户如zookeeper, kafka等。

安装Zookeeper

首先下载解压zookeeper,选择合适的镜像站点以加快下载速度。
我们可以将zookeeper加到系统服务中，增加一个/etc/init.d/zookeeper文件。

cd /opt
wget  http://apache.fayea.com/apache-mirror/zookeeper/zookeeper-3.4.6/zookeeper-3.4.6.tar.gz
tar zxvf zookeeper-3.4.6.tar.gz
vi /etc/init.d/zookeeper

将https://raw.githubusercontent.com/apache/zookeeper/trunk/src/packages/rpm/init.d/zookeeper文件的内容拷贝到这个文件，修改其中的运行zookeeper的用户以及zookeeper的文件夹位置。

......
start() {
  echo -n $"Starting $desc (zookeeper): "
  daemon --user root /opt/zookeeper-3.4.6/zkServer.sh start
  RETVAL=$?
  echo
  [ $RETVAL -eq 0 ] && touch /var/lock/subsys/zookeeper
  return $RETVAL
}

stop() {
  echo -n $"Stopping $desc (zookeeper): "
  daemon --user root /opt/zookeeper-3.4.6/zkServer.sh stop
  RETVAL=$?
  sleep 5
  echo
  [ $RETVAL -eq 0 ] && rm -f /var/lock/subsys/zookeeper $PIDFILE
}
......

1 2	chmod 755 /etc/init.d/zookeeper service zookeeper start

如果你不想加到服务，也可以直接运行zookeeper。

1	/opt/zookeeper-3.4.6/zkServer.sh start

安装Kafka

从合适的镜像站点下载最新的kafka并解压。

1
2
3

wget http://apache.01link.hk/kafka/0.8.2-beta/kafka_2.9.1-0.8.2-beta.tgz
tar zxvf kafka_2.9.1-0.8.2-beta.tgz
cd kafka_2.9.1-0.8.2-beta

启动Kafka:

1	bin/kafka-server-start.sh config/server.properties

创建一个test的topic：

1	bin/kafka-topics.sh --create --zookeeper localhost:2181 --replication-factor 1 --partitions 1 --topic test

可以利用kafka的命令启动一个生产者和消费者试验一下：

1
2
3

> bin/kafka-console-producer.sh --broker-list localhost:9092 --topic test
This is a message
This is another message

1
2
3

> bin/kafka-console-consumer.sh --zookeeper localhost:2181 --topic test --from-beginning
This is a message
This is another message

更多的介绍可以查看我翻译整理的 Kafka快速入门

创建一个Spring项目

以上的准备环境完成，让我们开始创建一个项目。
以前我写过一篇简单介绍: Spring 集成 Kafka.
spring-integration-kafka这个官方框架我就不介绍了。我们主要使用它做集成。

首先我们先看一下使用Kafka自己的Producer/Consumer API发送/接收消息的例子。

使用Producer API发送消息到Kafka

OK，现在我们先看一个使用Kafka 自己的producer API发送消息的例子：

public class NativeProducer {

	public static void main(String[] args) {
		String topic= "test";
		long events = 100;
        Random rand = new Random();
 
        Properties props = new Properties();
        props.put("metadata.broker.list", "localhost:9092");
        props.put("serializer.class", "kafka.serializer.StringEncoder");
        props.put("request.required.acks", "1");
 
        ProducerConfig config = new ProducerConfig(props);
 
        Producer<String, String> producer = new Producer<String, String>(config);
 
        for (long nEvents = 0; nEvents < events; nEvents++) {                
               String msg = "NativeMessage-" + rand.nextInt() ; 
               KeyedMessage<String, String> data = new KeyedMessage<String, String>(topic, nEvents + "", msg);
               producer.send(data);
        }
        producer.close();

	}
}

这个例子中首先初始化Producer对象，指定相应的broker和serializer，然后发送100个字符串消息给Kafka。

运行mvn package编译代码，执行查看结果：

1	java -cp target/lib/*:target/spring-kafka-demo-0.2.0-SNAPSHOT.jar com.colobu.spring_kafka_demo.NativeProducer

上面的消费者控制台窗口会打印出收到的消息:

......
NativeMessage--1645592376
NativeMessage-534168193
NativeMessage--1899432197
NativeMessage-1642480773
NativeMessage--911267171
NativeMessage-251458151
NativeMessage--55710397
NativeMessage-455515562
NativeMessage-1108982916
NativeMessage--1710296834
NativeMessage-2102648373
NativeMessage-499979365
NativeMessage--1200107003
NativeMessage-1184836299
NativeMessage--1161123005
NativeMessage-912582115
NativeMessage--1557863408
NativeMessage--1036456356
......

使用Kafka High Level API接收消息

用High level Consumer API接收消息，

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.ConsumerConfig;
import kafka.consumer.ConsumerIterator;
import kafka.consumer.KafkaStream;
import kafka.javaapi.consumer.ConsumerConnector;

public class NativeConsumer {
	private final ConsumerConnector consumer;
	private final String topic;
	private ExecutorService executor;

	public NativeConsumer(String a_zookeeper, String a_groupId, String a_topic) {
		consumer = kafka.consumer.Consumer.createJavaConsumerConnector(createConsumerConfig(a_zookeeper, a_groupId));
		this.topic = a_topic;
	}

	public void shutdown() {
		if (consumer != null)
			consumer.shutdown();
		if (executor != null)
			executor.shutdown();
	}

	public void run(int a_numThreads) {
		Map<String, Integer> topicCountMap = new HashMap<String, Integer>();
		topicCountMap.put(topic, new Integer(a_numThreads));
		Map<String, List<KafkaStream<byte[], byte[]>>> consumerMap = consumer.createMessageStreams(topicCountMap);
		List<KafkaStream<byte[], byte[]>> streams = consumerMap.get(topic);

		// now launch all the threads
		//
		executor = Executors.newFixedThreadPool(a_numThreads);

		// now create an object to consume the messages
		//
		int threadNumber = 0;
		for (final KafkaStream stream : streams) {
			executor.submit(new ConsumerTest(stream, threadNumber));
			threadNumber++;
		}
	}

	private static ConsumerConfig createConsumerConfig(String a_zookeeper, String a_groupId) {
		Properties props = new Properties();
		props.put("zookeeper.connect", a_zookeeper);
		props.put("group.id", a_groupId);
		props.put("zookeeper.session.timeout.ms", "400");
		props.put("zookeeper.sync.time.ms", "200");
		props.put("auto.commit.interval.ms", "1000");

		return new ConsumerConfig(props);
	}

	public static void main(String[] args) {
		String zooKeeper = "localhost:2181";
		String groupId = "mygroup";
		String topic = "test";
		int threads = 1;

		NativeConsumer example = new NativeConsumer(zooKeeper, groupId, topic);
		example.run(threads);

		try {
			Thread.sleep(10000);
		} catch (InterruptedException ie) {

		}
		//example.shutdown();
	}
}

class ConsumerTest implements Runnable {
    private KafkaStream m_stream;
    private int m_threadNumber;
 
    public ConsumerTest(KafkaStream a_stream, int a_threadNumber) {
        m_threadNumber = a_threadNumber;
        m_stream = a_stream;
    }
 
    public void run() {
        ConsumerIterator<byte[], byte[]> it = m_stream.iterator();
        while (it.hasNext())
            System.out.println("Thread " + m_threadNumber + ": " + new String(it.next().message()));
        System.out.println("Shutting down Thread: " + m_threadNumber);
    }
}

在生产者控制台输入几条消息，可以看到运行这个例子的控制台可以将这些消息打印出来。

教程的代码中还包括一个使用Simple Consumer API接收消息的例子。因为spring-integration-kafka不支持这种API，这里也不列出对比代码了。

使用spring-integration-kafka发送消息

Outbound Channel Adapter用来发送消息到Kafka。消息从Spring Integration Channel中读取。你可以在Spring application context指定这个channel。
一旦配置好这个Channel，就可以利用这个Channel往Kafka发消息。明显地，Spring Integration特定的消息发送给这个Adaptor，然后发送前在内部被转为Kafka消息。当前的版本要求你必须指定消息key和topic作为头部数据 (header)，消息作为有载荷(payload)。
例如

final MessageChannel channel = ctx.getBean("inputToKafka", MessageChannel.class);

channel.send(
        MessageBuilder.withPayload(payload) //设置有效载荷
                .setHeader("messageKey", "key") //指定key
                .setHeader("topic", "test").build()); /指定topic/

实际代码如下：

import java.util.Random;

import org.springframework.context.support.ClassPathXmlApplicationContext;
import org.springframework.integration.support.MessageBuilder;
import org.springframework.messaging.MessageChannel;

public class Producer {
	private static final String CONFIG = "/context.xml";
	private static Random rand = new Random();

	public static void main(String[] args) {
		final ClassPathXmlApplicationContext ctx = new ClassPathXmlApplicationContext(CONFIG, Producer.class);
		ctx.start();

		final MessageChannel channel = ctx.getBean("inputToKafka", MessageChannel.class);

		for (int i = 0; i < 100; i++) {
			channel.send(MessageBuilder.withPayload("Message-" + rand.nextInt()).setHeader("messageKey", String.valueOf(i)).setHeader("topic", "test").build());
		}

		try {
			Thread.sleep(100000);
		} catch (InterruptedException e) {
			e.printStackTrace();
		}
		ctx.close();
	}
}

Spring 配置文件：

<?xml version="1.0" encoding="UTF-8"?>
<beans xmlns="http://www.springframework.org/schema/beans"
       xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
       xmlns:int="http://www.springframework.org/schema/integration"
       xmlns:int-kafka="http://www.springframework.org/schema/integration/kafka"
       xmlns:task="http://www.springframework.org/schema/task"
       xsi:schemaLocation="http://www.springframework.org/schema/integration/kafka http://www.springframework.org/schema/integration/kafka/spring-integration-kafka.xsd
		http://www.springframework.org/schema/integration http://www.springframework.org/schema/integration/spring-integration.xsd
		http://www.springframework.org/schema/beans http://www.springframework.org/schema/beans/spring-beans.xsd
		http://www.springframework.org/schema/task http://www.springframework.org/schema/task/spring-task.xsd">

    <int:channel id="inputToKafka">
        <int:queue/>
    </int:channel>

    <int-kafka:outbound-channel-adapter id="kafkaOutboundChannelAdapter"
                                        kafka-producer-context-ref="kafkaProducerContext"
                                        auto-startup="false"
                                        channel="inputToKafka"
                                        order="3"
            >
        <int:poller fixed-delay="1000" time-unit="MILLISECONDS" receive-timeout="0" task-executor="taskExecutor"/>
    </int-kafka:outbound-channel-adapter>

    <task:executor id="taskExecutor" pool-size="5" keep-alive="120" queue-capacity="500"/>

	<bean id="producerProperties"
		class="org.springframework.beans.factory.config.PropertiesFactoryBean">
		<property name="properties">
			<props>
				<prop key="topic.metadata.refresh.interval.ms">3600000</prop>
				<prop key="message.send.max.retries">5</prop>
				<prop key="serializer.class">kafka.serializer.StringEncoder</prop>
				<prop key="request.required.acks">1</prop>
			</props>
		</property>
	</bean>

	<int-kafka:producer-context id="kafkaProducerContext"
		producer-properties="producerProperties">
		<int-kafka:producer-configurations>
			<int-kafka:producer-configuration broker-list="localhost:9092"
                       topic="test"
                       compression-codec="default"/>
		</int-kafka:producer-configurations>
	</int-kafka:producer-context>
</beans>

int:channel是配置Spring Integration Channel, 此channel基于queue。
int-kafka:outbound-channel-adapter是outbound-channel-adapter对象，内部使用一个线程池处理消息。关键是kafka-producer-context-ref。
int-kafka:producer-context配置producer列表，要处理的topic，这些Producer最终要转换成Kafka的Producer。

producer的配置参数如下：

broker-list             List of comma separated brokers that this producer connects to
topic                   Topic name or Java regex pattern of topic name
compression-codec       Compression method to be used. Default is no compression. Supported compression codec are gzip and snappy.
                        Anything else would result in no compression
value-encoder           Serializer to be used for encoding messages.
key-encoder             Serializer to be used for encoding the partition key
key-class-type          Type of the key class. This will be ignored if no key-encoder is provided
value-class-type        Type of the value class. This will be ignored if no value-encoder is provided.
partitioner             Custom implementation of a Kafka Partitioner interface.
async                   True/False - default is false. Setting this to true would make the Kafka producer to use
                        an async producer
batch-num-messages      Number of messages to batch at the producer. If async is false, then this has no effect.

value-encoder 和key-encoder可以是其它实现了Kafka Encoder接口的Bean。同样partitioner也是实现了Kafka的Partitioner接口的Bean。
一个Encoder的例子：

1
2
3

<bean id="kafkaEncoder" class="org.springframework.integration.kafka.serializer.avro.AvroSpecificDatumBackedKafkaEncoder">
    <constructor-arg value="com.company.AvroGeneratedSpecificRecord" />
</bean>

Spring Integration Kafka 也提供了个基于Avro的Encoder。 Avro也是Apache的一个项目，在大数据处理时也是一个常用的序列化框架。
不指定Encoder将使用Kafka缺省的Encoder (kafka.serializer.DefaultEncoder, byte[] --> same byte[])。

producerProperties可以用来设置配置属性进行调优。配置属性列表请参考 http://kafka.apache.org/documentation.html#producerconfigs

使用spring-integration-kafka接收消息

同样的原理实现一个消费者：

package com.colobu.spring_kafka_demo;

import java.util.Collection;
import java.util.HashMap;
import java.util.Iterator;
import java.util.List;
import java.util.Map;
import java.util.Random;
import java.util.Set;
import java.util.concurrent.ConcurrentHashMap;

import org.slf4j.LoggerFactory;
import org.springframework.context.support.ClassPathXmlApplicationContext;
import org.springframework.integration.channel.QueueChannel;
import org.springframework.messaging.Message;

import ch.qos.logback.classic.Level;

public class Consumer {
	private static final String CONFIG = "/consumer_context.xml";
	private static Random rand = new Random();

	@SuppressWarnings({ "unchecked", "unchecked", "rawtypes" })
	public static void main(String[] args) {
		ch.qos.logback.classic.Logger rootLogger = (ch.qos.logback.classic.Logger)LoggerFactory.getLogger(ch.qos.logback.classic.Logger.ROOT_LOGGER_NAME);
		rootLogger.setLevel(Level.toLevel("info"));
		
		final ClassPathXmlApplicationContext ctx = new ClassPathXmlApplicationContext(CONFIG, Consumer.class);
		ctx.start();

		final QueueChannel channel = ctx.getBean("inputFromKafka", QueueChannel.class);
		Message msg;		
		while((msg = channel.receive()) != null) {
			HashMap map = (HashMap)msg.getPayload();
			Set<Map.Entry> set = map.entrySet();
			for (Map.Entry entry : set) {
				String topic = (String)entry.getKey();
				System.out.println("Topic:" + topic);
				ConcurrentHashMap<Integer,List<byte[]>> messages = (ConcurrentHashMap<Integer,List<byte[]>>)entry.getValue();
				Collection<List<byte[]>> values = messages.values();
				for (Iterator<List<byte[]>> iterator = values.iterator(); iterator.hasNext();) {
					List<byte[]> list = iterator.next();
					for (byte[] object : list) {
						String message = new String(object);
						System.out.println("\tMessage: " + message);
					}
					
				}
			
			}
			
		}
		
		try {
			Thread.sleep(100000);
		} catch (InterruptedException e) {
			e.printStackTrace();
		}
		ctx.close();
	}
}

Spring的配置文件如下：

<?xml version="1.0" encoding="UTF-8"?>
<beans xmlns="http://www.springframework.org/schema/beans"
	xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:int="http://www.springframework.org/schema/integration"
	xmlns:int-kafka="http://www.springframework.org/schema/integration/kafka"
	xmlns:task="http://www.springframework.org/schema/task"
	xsi:schemaLocation="http://www.springframework.org/schema/integration/kafka http://www.springframework.org/schema/integration/kafka/spring-integration-kafka.xsd
		http://www.springframework.org/schema/integration http://www.springframework.org/schema/integration/spring-integration.xsd
		http://www.springframework.org/schema/beans http://www.springframework.org/schema/beans/spring-beans.xsd
		http://www.springframework.org/schema/task http://www.springframework.org/schema/task/spring-task.xsd">

    <int:channel id="inputFromKafka">
        <int:queue/>
    </int:channel>
    
	<int-kafka:inbound-channel-adapter
		id="kafkaInboundChannelAdapter" kafka-consumer-context-ref="consumerContext"
		auto-startup="false" channel="inputFromKafka">
		<int:poller fixed-delay="10" time-unit="MILLISECONDS"
			max-messages-per-poll="5" />
	</int-kafka:inbound-channel-adapter>

	<bean id="consumerProperties"
		class="org.springframework.beans.factory.config.PropertiesFactoryBean">
		<property name="properties">
			<props>
				<prop key="auto.offset.reset">smallest</prop>
				<prop key="socket.receive.buffer.bytes">10485760</prop> <!-- 10M -->
				<prop key="fetch.message.max.bytes">5242880</prop>
				<prop key="auto.commit.interval.ms">1000</prop>
			</props>
		</property>
	</bean>

	<int-kafka:consumer-context id="consumerContext"
		consumer-timeout="4000" zookeeper-connect="zookeeperConnect" consumer-properties="consumerProperties">
		<int-kafka:consumer-configurations>
			<int-kafka:consumer-configuration
				group-id="mygroup" max-messages="5000">
				<int-kafka:topic id="test" streams="4" />
			</int-kafka:consumer-configuration>
			<!-- <int-kafka:consumer-configuration group-id="default3" value-decoder="kafkaSpecificDecoder" 
				key-decoder="kafkaReflectionDecoder" max-messages="10"> <int-kafka:topic-filter 
				pattern="regextopic.*" streams="4" exclude="false" /> </int-kafka:consumer-configuration> -->
		</int-kafka:consumer-configurations>
	</int-kafka:consumer-context>

	<int-kafka:zookeeper-connect id="zookeeperConnect"
		zk-connect="localhost:2181" zk-connection-timeout="6000"
		zk-session-timeout="400" zk-sync-time="200" />
</beans>

这个配置和Producer类似，同样声明一个channel, 定义inbound-channel-adapter, 它引用Bean kafka-consumer-context,
kafka-consumer-context定义了消费者的列表。 consumer-configuration还提供了topic-filter，使用正则表达式建立白名单或者黑名单(exclude属性)。

消费者上下文还需要zookeeper-connect。

由于spring-integration-kafka只实现了high level Consumer API,这也就意味着你不可能回滚重新查看以前的消息，因为high level API不提供offset管理。

注意Channel中得到的有效负载的类型是：
Map<String, Map<Integer, List<Object>>>,
这个Map的key是topic, 值还是另外的Map对象。
这个值Map的key值是分区号，value值是消息列表。在本例中由于消息是字符串，转换成了byte[]数组。

这种复杂的结构是由于Kafka的设计造成的。 Kafka保证对于同一个topic的同一个分区的消息是严格有序的。所有这种数据结构可以提供有序的消息。