概述
kafka中有个 micro batch 的概念 ，为了提高Producer 发送的性能。
不同于RocketMQ 提供了一个可以批量发送多条消息的 API 。 Kafka 的做法是：提供了一个 RecordAccumulator 消息收集器，将发送给相同 Topic 的相同 Partition 分区的消息们，缓冲一下，当满足条件时候，一次性批量将缓冲的消息提交给 Kafka Broker 。
参数设置
https://kafka.apache.org/24/documentation.html#producerconfigs
主要涉及的参数 ，三个条件，满足任一即会批量发送：
- batch-size ：超过收集的消息数量的最大量。默认16KB
- buffer-memory ：超过收集的消息占用的最大内存 , 默认32M
- linger.ms ：超过收集的时间的最大等待时长，单位：毫秒。
Code
POM依赖
<dependencies> <dependency> <groupId>org.springframework.boot</groupId> <artifactId>spring-boot-starter-web</artifactId> </dependency> <!-- 引入 Spring-Kafka 依赖 --> <dependency> <groupId>org.springframework.kafka</groupId> <artifactId>spring-kafka</artifactId> </dependency> <dependency> <groupId>org.springframework.boot</groupId> <artifactId>spring-boot-starter-test</artifactId> <scope>test</scope> </dependency> <dependency> <groupId>junit</groupId> <artifactId>junit</artifactId> <scope>test</scope> </dependency> </dependencies>
配置文件
spring: # Kafka 配置项，对应 KafkaProperties 配置类 kafka: bootstrap-servers: 192.168.126.140:9092 # 指定 Kafka Broker 地址，可以设置多个，以逗号分隔 # Kafka Producer 配置项 producer: acks: 1 # 0-不应答。1-leader 应答。all-所有 leader 和 follower 应答。 retries: 3 # 发送失败时，重试发送的次数 key-serializer: org.apache.kafka.common.serialization.StringSerializer # 消息的 key 的序列化 value-serializer: org.springframework.kafka.support.serializer.JsonSerializer # 消息的 value 的序列化 batch-size: 16384 # 每次批量发送消息的最大数量 单位 字节 默认 16K buffer-memory: 33554432 # 每次批量发送消息的最大内存 单位 字节 默认 32M properties: linger: ms: 10000 # 批处理延迟时间上限。[实际不会配这么长，这里用于测速]这里配置为 10 * 1000 ms 过后，不管是否消息数量是否到达 batch-size 或者消息大小到达 buffer-memory 后，都直接发送一次请求。 # Kafka Consumer 配置项 consumer: auto-offset-reset: earliest # 设置消费者分组最初的消费进度为 earliest key-deserializer: org.apache.kafka.common.serialization.StringDeserializer value-deserializer: org.springframework.kafka.support.serializer.JsonDeserializer properties: spring: json: trusted: packages: com.artisan.springkafka.domain # Kafka Consumer Listener 监听器配置 listener: missing-topics-fatal: false # 消费监听接口监听的主题不存在时，默认会报错。所以通过设置为 false ，解决报错 logging: level: org: springframework: kafka: ERROR # spring-kafka apache: kafka: ERROR # kafka
生产者
package com.artisan.springkafka.producer; import com.artisan.springkafka.constants.TOPIC; import com.artisan.springkafka.domain.MessageMock; import org.springframework.beans.factory.annotation.Autowired; import org.springframework.kafka.core.KafkaTemplate; import org.springframework.kafka.support.SendResult; import org.springframework.stereotype.Component; import org.springframework.util.concurrent.ListenableFuture; import java.util.Random; import java.util.concurrent.ExecutionException; /** * @author 小工匠 * @version 1.0 * @description: TODO * @date 2021/2/17 22:25 * @mark: show me the code , change the world */ @Component public class ArtisanProducerMock { @Autowired private KafkaTemplate<Object,Object> kafkaTemplate ; /** * 同步发送 * @return * @throws ExecutionException * @throws InterruptedException */ public SendResult sendMsgSync() throws ExecutionException, InterruptedException { // 模拟发送的消息 Integer id = new Random().nextInt(100); MessageMock messageMock = new MessageMock(id,"artisanTestMessage-" + id); // 同步等待 return kafkaTemplate.send(TOPIC.TOPIC, messageMock).get(); } public ListenableFuture<SendResult<Object, Object>> sendMsgASync() throws ExecutionException, InterruptedException { // 模拟发送的消息 Integer id = new Random().nextInt(100); MessageMock messageMock = new MessageMock(id,"messageSendByAsync-" + id); // 异步发送消息 ListenableFuture<SendResult<Object, Object>> result = kafkaTemplate.send(TOPIC.TOPIC, messageMock); return result ; } }
消费者
package com.artisan.springkafka.consumer; import com.artisan.springkafka.domain.MessageMock; import com.artisan.springkafka.constants.TOPIC; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import org.springframework.kafka.annotation.KafkaListener; import org.springframework.stereotype.Component; /** * @author 小工匠 * @version 1.0 * @description: TODO * @date 2021/2/17 22:33 * @mark: show me the code , change the world */ @Component public class ArtisanCosumerMock { private Logger logger = LoggerFactory.getLogger(getClass()); private static final String CONSUMER_GROUP_PREFIX = "MOCK-A" ; @KafkaListener(topics = TOPIC.TOPIC ,groupId = CONSUMER_GROUP_PREFIX + TOPIC.TOPIC) public void onMessage(MessageMock messageMock){ logger.info("【接受到消息][线程:{} 消息内容：{}]", Thread.currentThread().getName(), messageMock); } }
package com.artisan.springkafka.consumer; import com.artisan.springkafka.domain.MessageMock; import com.artisan.springkafka.constants.TOPIC; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import org.springframework.kafka.annotation.KafkaListener; import org.springframework.stereotype.Component; /** * @author 小工匠 * @version 1.0 * @description: TODO * @date 2021/2/17 22:33 * @mark: show me the code , change the world */ @Component public class ArtisanCosumerMockDiffConsumeGroup { private Logger logger = LoggerFactory.getLogger(getClass()); private static final String CONSUMER_GROUP_PREFIX = "MOCK-B" ; @KafkaListener(topics = TOPIC.TOPIC ,groupId = CONSUMER_GROUP_PREFIX + TOPIC.TOPIC) public void onMessage(MessageMock messageMock){ logger.info("【接受到消息][线程:{} 消息内容：{}]", Thread.currentThread().getName(), messageMock); } }
单元测试
package com.artisan.springkafka.produceTest; import com.artisan.springkafka.SpringkafkaApplication; import com.artisan.springkafka.producer.ArtisanProducerMock; import org.junit.Test; import org.junit.runner.RunWith; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import org.springframework.beans.factory.annotation.Autowired; import org.springframework.boot.test.context.SpringBootTest; import org.springframework.kafka.support.SendResult; import org.springframework.test.context.junit4.SpringRunner; import org.springframework.util.concurrent.ListenableFuture; import org.springframework.util.concurrent.ListenableFutureCallback; import java.util.concurrent.CountDownLatch; import java.util.concurrent.ExecutionException; import java.util.concurrent.TimeUnit; /** * @author 小工匠 * * @version 1.0 * @description: TODO * @date 2021/2/17 22:40 * @mark: show me the code , change the world */ @RunWith(SpringRunner.class) @SpringBootTest(classes = SpringkafkaApplication.class) public class ProduceMockTest { private Logger logger = LoggerFactory.getLogger(getClass()); @Autowired private ArtisanProducerMock artisanProducerMock; @Test public void testAsynSend() throws ExecutionException, InterruptedException { logger.info("开始发送"); for (int i = 0; i < 2; i++) { artisanProducerMock.sendMsgASync().addCallback(new ListenableFutureCallback<SendResult<Object, Object>>() { @Override public void onFailure(Throwable throwable) { logger.info(" 发送异常{}]]", throwable); } @Override public void onSuccess(SendResult<Object, Object> objectObjectSendResult) { logger.info("回调结果 Result = topic:[{}] , partition:[{}], offset:[{}]", objectObjectSendResult.getRecordMetadata().topic(), objectObjectSendResult.getRecordMetadata().partition(), objectObjectSendResult.getRecordMetadata().offset()); } }); // 发送2次 每次间隔5秒， 凑够我们配置的 linger: ms: 10000 TimeUnit.SECONDS.sleep(5); } // 阻塞等待，保证消费 new CountDownLatch(1).await(); } }
异步发送2条消息，每次发送消息之间， sleep 5 秒，以便达到配置的 linger.ms 最大等待时长10秒。
测试结果
2021-02-18 10:58:53.360 INFO 24736 --- [ main] c.a.s.produceTest.ProduceMockTest : 开始发送 2021-02-18 10:59:03.555 INFO 24736 --- [ad | producer-1] c.a.s.produceTest.ProduceMockTest : 回调结果 Result = topic:[MOCK_TOPIC] , partition:[0], offset:[30] 2021-02-18 10:59:03.556 INFO 24736 --- [ad | producer-1] c.a.s.produceTest.ProduceMockTest : 回调结果 Result = topic:[MOCK_TOPIC] , partition:[0], offset:[31] 2021-02-18 10:59:03.595 INFO 24736 --- [ntainer#0-0-C-1] c.a.s.consumer.ArtisanCosumerMock : 【接受到消息][线程:org.springframework.kafka.KafkaListenerEndpointContainer#0-0-C-1 消息内容：MessageMock{id=6, name='messageSendByAsync-6'}] 2021-02-18 10:59:03.595 INFO 24736 --- [ntainer#1-0-C-1] a.s.c.ArtisanCosumerMockDiffConsumeGroup : 【接受到消息][线程:org.springframework.kafka.KafkaListenerEndpointContainer#1-0-C-1 消息内容：MessageMock{id=6, name='messageSendByAsync-6'}] 2021-02-18 10:59:03.595 INFO 24736 --- [ntainer#1-0-C-1] a.s.c.ArtisanCosumerMockDiffConsumeGroup : 【接受到消息][线程:org.springframework.kafka.KafkaListenerEndpointContainer#1-0-C-1 消息内容：MessageMock{id=94, name='messageSendByAsync-94'}] 2021-02-18 10:59:03.595 INFO 24736 --- [ntainer#0-0-C-1] c.a.s.consumer.ArtisanCosumerMock : 【接受到消息][线程:org.springframework.kafka.KafkaListenerEndpointContainer#0-0-C-1 消息内容：MessageMock{id=94, name='messageSendByAsync-94'}]
10 秒后，满足批量消息的最大等待时长，所以 2 条消息被 Producer 批量发送。同时我们配置的是 acks=1 ，需要等待发送成功后，才会回调 ListenableFutureCallback 的方法。
当然了，我们这里都是为了测试，设置的这么长的间隔，实际中需要根据具体的业务场景设置一个合理的值。
源码地址
https://github.com/yangshangwei/boot2/tree/master/springkafkaBatchSend
