Java使用Kafka的方法
本篇内容主要讲解“Java使用Kafka的方法”,感兴趣的朋友不妨来看看。本文介绍的方法操作简单快捷,实用性强。下面就让小编来带大家学习“Java使用Kafka的方法”吧!
创新互联公司于2013年开始,是专业互联网技术服务公司,拥有项目网站制作、成都网站制作网站策划,项目实施与项目整合能力。我们以让每一个梦想脱颖而出为使命,1280元梓潼做网站,已为上家服务,为梓潼各地企业和个人服务,联系电话:18980820575
1、maven依赖
org.apache.kafka kafka-clients 0.11.0.0
2、Producer
2.1、producer发送消息
import java.util.Properties; import org.apache.kafka.clients.producer.KafkaProducer; import org.apache.kafka.clients.producer.Producer; import org.apache.kafka.clients.producer.ProducerRecord; /** * @author Thomas * @Description:最简单的kafka producer * @date 22:18 2019-7-5 */ public class ProducerDemo { public static void main(String[] args) { Properties properties =new Properties(); //zookeeper服务器集群地址,用逗号隔开 properties.put("bootstrap.servers", "172.16.0.218:9092,172.16.0.219:9092,172.16.0.217:9092"); properties.put("acks", "all"); properties.put("retries", 0); properties.put("batch.size", 16384); properties.put("linger.ms", 1); properties.put("buffer.memory", 33554432); properties.put("key.serializer", "org.apache.kafka.common.serialization.StringSerializer"); properties.put("value.serializer", "org.apache.kafka.common.serialization.StringSerializer"); //自定义producer拦截器 properties.put("interceptor.classes", "com.lt.kafka.producer.MyProducerInterceptor"); //自定义消息路由规则(消息发送到哪一个Partition中) //properties.put("partitioner.class", "com.lt.kafka.producer.MyPartition"); Producerproducer = null; try { producer = new KafkaProducer (properties); for (int i = 20; i < 40; i++) { String msg = "This is Message:" + i; /** * kafkaproducer中会同时调用自己的callback的onCompletion方法和producerIntercepter的onAcknowledgement方法。 * 关键源码:Callback interceptCallback = this.interceptors == null * callback : new InterceptorCallback<>(callback, * this.interceptors, tp); */ producer.send(new ProducerRecord ("leixiang", msg),new MyCallback()); } } catch (Exception e) { e.printStackTrace(); } finally { if(producer!=null) producer.close(); } } }
2.2、自定义producer拦截器
import java.util.Map; import org.apache.kafka.clients.producer.ProducerInterceptor; import org.apache.kafka.clients.producer.ProducerRecord; import org.apache.kafka.clients.producer.RecordMetadata; /** * @author Thomas * @Description:自定义producer拦截器 * @date 22:21 2019-7-5 */ public class MyProducerInterceptor implements ProducerInterceptor{ /** * 打印配置相关信息 */ public void configure(Map configs) { // TODO Auto-generated method stub System.out.println(configs.toString()); } /** * producer发送信息拦截方法 */ public ProducerRecord onSend(ProducerRecord record) { System.out.println("拦截处理前============="); String topic=record.topic(); String value=record.value(); System.out.println("拦截处理前的消息====:"+value); ProducerRecord record2=new ProducerRecord (topic, value+" (intercepted)"); System.out.println("拦截处理后的消息:"+record2.value()); System.out.println("拦截处理后==============="); return record2; } /** * 消息确认回调函数,和callback的onCompletion方法相似。 * 在kafkaProducer中,如果都设置,两者都会调用。 */ public void onAcknowledgement(RecordMetadata metadata, Exception exception) { if (metadata != null) System.out.println("MyProducerInterceptor onAcknowledgement:RecordMetadata=" + metadata.toString()); if (exception != null) exception.printStackTrace(); } /** * interceptor关闭回调 */ public void close() { System.out.println("MyProducerInterceptor is closed!"); } }
2.3、自定义消息路由规则
自定义路由规则,可以根据自己的需要定义消息发送到哪个分区。自定义路由规则需要实现Partitioner。
import java.util.Map; import org.apache.kafka.clients.producer.Partitioner; import org.apache.kafka.common.Cluster; /** * @author Thomas * @Description: * @date 22:24 2019-7-5 */ public class MyPartition implements Partitioner { public void configure(Maparg0) { // TODO Auto-generated method stub } public void close() { // TODO Auto-generated method stub } public int partition(String arg0, Object arg1, byte[] arg2, Object arg3, byte[] arg4, Cluster arg5) { // TODO Auto-generated method stub return 0; } }
3、Consumer
3.1、自动提交
import java.util.Arrays; import java.util.Properties; import org.apache.kafka.clients.consumer.ConsumerRecord; import org.apache.kafka.clients.consumer.ConsumerRecords; import org.apache.kafka.clients.consumer.KafkaConsumer; /** * @author Thomas * @Description: * @date 22:26 2019-7-5 */ public class AutoCommitConsumerDemo { public static void main(String[] args) { Properties props = new Properties(); props.put("bootstrap.servers", "172.16.0.218:9092,172.16.0.219:9092,172.16.0.217:9092"); props.put("group.id", "leixiang"); props.put("enable.auto.commit", "true"); //想要读取之前的数据,必须加上 //props.put("auto.offset.reset", "earliest"); /* 自动确认offset的时间间隔 */ props.put("auto.commit.interval.ms", "1000"); /* * 一旦consumer和kakfa集群建立连接, * consumer会以心跳的方式来高速集群自己还活着, * 如果session.timeout.ms 内心跳未到达服务器,服务器认为心跳丢失,会做rebalence */ props.put("session.timeout.ms", "30000"); props.put("key.deserializer", "org.apache.kafka.common.serialization.StringDeserializer"); props.put("value.deserializer", "org.apache.kafka.common.serialization.StringDeserializer"); //配置自定义的拦截器,可以在拦截器中引入第三方插件实现日志记录等功能。 //props.put("interceptor.classes", "com.lt.kafka.consumer.MyConsumerInterceptor"); @SuppressWarnings("resource") KafkaConsumerconsumer = new KafkaConsumer (props); try { /* 消费者订阅的topic, 可同时订阅多个 ,用逗号隔开*/ consumer.subscribe(Arrays.asList("leixiang")); while (true) { //轮询数据。如果缓冲区中没有数据,轮询等待的时间为毫秒。如果0,立即返回缓冲区中可用的任何记录,则返回空 ConsumerRecords records = consumer.poll(100); for (ConsumerRecord record : records) System.out.printf("offset = %d, key = %s, value = %s%n", record.offset(), record.key(), record.value()); } } catch (Exception e) { // TODO: handle exception e.printStackTrace(); } } }
3.2、手动提交
import java.util.Arrays; import java.util.Properties; import org.apache.kafka.clients.consumer.ConsumerRecord; import org.apache.kafka.clients.consumer.ConsumerRecords; import org.apache.kafka.clients.consumer.KafkaConsumer; /** * @author Thomas * @Description: * @date 22:28 2019-7-5 */ public class ManualCommitConsumerDemo { public static void main(String[] args) { Properties props = new Properties(); props.put("bootstrap.servers", "172.16.0.218:9092,172.16.0.219:9092,172.16.0.217:9092"); props.put("group.id", "leixiang"); props.put("enable.auto.commit", "false");//手动确认 /* 自动确认offset的时间间隔 */ props.put("auto.commit.interval.ms", "1000"); props.put("auto.offset.reset", "earliest");//想要读取之前的数据,必须加上 /* * 一旦consumer和kakfa集群建立连接, * consumer会以心跳的方式来高速集群自己还活着, * 如果session.timeout.ms 内心跳未到达服务器,服务器认为心跳丢失,会做rebalence */ props.put("session.timeout.ms", "30000"); props.put("key.deserializer", "org.apache.kafka.common.serialization.StringDeserializer"); props.put("value.deserializer", "org.apache.kafka.common.serialization.StringDeserializer"); //配置自定义的拦截器,可以在拦截器中引入第三方插件实现日志记录等功能。 props.put("interceptor.classes", "com.lt.kafka.consumer.MyConsumerInterceptor"); KafkaConsumerconsumer = new KafkaConsumer (props); /* 消费者订阅的topic, 可同时订阅多个 ,用逗号隔开*/ consumer.subscribe(Arrays.asList("leixiang")); while (true) { ConsumerRecords records = consumer.poll(100); for (ConsumerRecord record : records) { //处理消息 saveMessage(record); //手动提交,并且设置Offset提交回调方法 //consumer.commitAsync(new MyOffsetCommitCallback()); consumer.commitAsync(); } } } public static void saveMessage(ConsumerRecord record){ System.out.printf("处理消息:offset = %d, key = %s, value = %s%n", record.offset(), record.key(), record.value()); } }
自定义Consumer拦截器
import java.util.Map; import org.apache.kafka.clients.consumer.ConsumerInterceptor; import org.apache.kafka.clients.consumer.ConsumerRecords; import org.apache.kafka.clients.consumer.OffsetAndMetadata; import org.apache.kafka.common.TopicPartition; /** * @author Thomas * @Description: * @date 22:29 2019-7-5 */ public class MyConsumerInterceptor implements ConsumerInterceptor{ public void configure(Map configs) { System.out.println("MyConsumerInterceptor configs>>>"+configs.toString()); } public ConsumerRecords onConsume(ConsumerRecords records) { System.out.println("onConsume"); return records; } public void onCommit(Map offsets) { System.out.println("onCommit"); } public void close() { System.out.println("MyConsumerInterceptor is closed!"); } }
自定义Offset提交回调方法
import java.util.Map; import org.apache.kafka.clients.consumer.OffsetAndMetadata; import org.apache.kafka.clients.consumer.OffsetCommitCallback; import org.apache.kafka.common.TopicPartition; /** * @author Thomas * @Description: * @date 22:31 2019-7-5 */ public class MyOffsetCommitCallback implements OffsetCommitCallback { public void onComplete(Mapoffsets, Exception exception) { if (offsets != null) System.out.println("offsets>>>" + offsets.toString()); if (exception != null) exception.printStackTrace(); } }
到此,相信大家对“Java使用Kafka的方法”有了更深的了解,不妨来实际操作一番吧!这里是创新互联网站,更多相关内容可以进入相关频道进行查询,关注我们,继续学习!
网页标题:Java使用Kafka的方法
网站路径:http://pwwzsj.com/article/jidhec.html