线上问题:SpringBoot中kafka消费者自动提交关闭无效

1 场景复现

关闭消费者自动提交offset，即enable-auto-commit: false，但是，消费者消费数据后，查询group-id的offset还是更新了。

配置profile：

spring:
  application:
    name: tutorial # 服务名称:注册中心server
  devtools:
    restart:
      enabled: true
  kafka:
    bootstrap-servers: 192.168.211.129:9092
    producer:
      retries: 0
      batch-size: 16384
      buffer-memory: 33554432
      key-serializer: org.apache.kafka.common.serialization.StringSerializer
      value-serializer: org.apache.kafka.common.serialization.StringSerializer
      properties:
        linger.ms: 1
    consumer:
      enable-auto-commit: false
      auto-commit-interval: 100ms
      key-deserializer: org.apache.kafka.common.serialization.StringDeserializer
      value-deserializer: org.apache.kafka.common.serialization.StringDeserializer
      properties:
        session.timeout.ms: 15000
      group-id: tutorial

消费者：

package com.monkey.tutorial.modules.kafka;

import org.apache.kafka.clients.consumer.ConsumerRecord;
import org.apache.kafka.clients.consumer.ConsumerRecords;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.kafka.annotation.KafkaListener;
import org.springframework.kafka.support.Acknowledgment;
import org.springframework.stereotype.Service;

/**
 * 我是描述信息.
 *
 * @author xindaqi
 * @date 2022-01-20 12:01
 */
@Service
public class Consumer {

    private static final Logger logger = LoggerFactory.getLogger(Consumer.class);

    @KafkaListener(topics = {"tutorial"}, groupId = "g1")
    public void consumer1(ConsumerRecord<String, String> consumerRecord) {
        String topic = consumerRecord.topic();
        String data = consumerRecord.value();
        logger.info(">>>>>>>>>>Kafka consumer1, topic:{}, data:{}", topic, data);
    }
}

2 原因

Kafka包中的消费者的自动提交不能单独使用，enable-auto-commit单独使用是无效的，即单独配置，最终都会自动提交offset。
当只配置消费者enable-auto-commit: false，不会生效，因为此时手动提交也为false，按照表达式：!this.isAnyManualAck && !this.autoCommit最终都会为true，从而自动提交。
方法：org.springframework.kafka.listener.KafkaMessageListenerContainer.ListenerConsumer#invokeBatchOnMessage

3 方案

消费者自动提交offset时，无需配置参数：enable-auto-commit；

spring:
  application:
    name: tutorial # 服务名称:注册中心server
  devtools:
    restart:
      enabled: true
  kafka:
    bootstrap-servers: 192.168.211.129:9092
    producer:
      retries: 0
      batch-size: 16384
      buffer-memory: 33554432
      key-serializer: org.apache.kafka.common.serialization.StringSerializer
      value-serializer: org.apache.kafka.common.serialization.StringSerializer
      properties:
        linger.ms: 1
    consumer:
      auto-commit-interval: 100ms
      key-deserializer: org.apache.kafka.common.serialization.StringDeserializer
      value-deserializer: org.apache.kafka.common.serialization.StringDeserializer
      properties:
        session.timeout.ms: 15000
      group-id: tutorial

消费者手动提交offset时，配置enable-auto-commit: false和ack-mode: manual。

spring:
  application:
    name: tutorial # 服务名称:注册中心server
  devtools:
    restart:
      enabled: true
  kafka:
    bootstrap-servers: 192.168.211.129:9092
    producer:
      retries: 0
      batch-size: 16384
      buffer-memory: 33554432
      key-serializer: org.apache.kafka.common.serialization.StringSerializer
      value-serializer: org.apache.kafka.common.serialization.StringSerializer
      properties:
        linger.ms: 1
    consumer:
      enable-auto-commit: false
      key-deserializer: org.apache.kafka.common.serialization.StringDeserializer
      value-deserializer: org.apache.kafka.common.serialization.StringDeserializer
      properties:
        session.timeout.ms: 15000
      group-id: tutorial
    listener:
      missing-topics-fatal: false
      type: batch
      ack-mode: manual

手动ACK，自定义业务逻辑，样例如下：

package com.monkey.tutorial.modules.kafka;

import org.apache.kafka.clients.consumer.ConsumerRecord;
import org.apache.kafka.clients.consumer.ConsumerRecords;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.kafka.annotation.KafkaListener;
import org.springframework.kafka.support.Acknowledgment;
import org.springframework.stereotype.Service;

/**
 * 我是描述信息.
 *
 * @author xindaqi
 * @date 2022-01-20 12:01
 */
@Service
public class Consumer {

    private static final Logger logger = LoggerFactory.getLogger(Consumer.class);

    @KafkaListener(topics = {"tutorial"}, groupId = "g1")
    public void consumer1(ConsumerRecords<String, String> consumerRecords, Acknowledgment acknowledgment) {
        for(ConsumerRecord<String, String> record : consumerRecords) {
            String topic = record.topic();
            String data = record.value();

            logger.info(">>>>>>>>>>Kafka consumer1, topic:{}, data:{}", topic, data);
            try {
            	// TODO 业务逻辑
                acknowledgment.acknowledge();
                logger.info(">>>>>>>>>>成功--Group g1 consumer手动上报offset");
            } catch(Exception ex) {
                logger.info(">>>>>>>>失败--Group g1 consumer手动上报offset");
            }
        }
    }
}

4 小结

Kafka包中的消费者的自动提交不能单独使用，enable-auto-commit单独使用是无效的，即单独配置，最终都会自动提交offset。
（1）enable-auto-commit是为手动提交设置的；
（2）Ack手动提交，配置消费者enable-auto-commit: false和手动提交ack-mode: manual，此时，在消费者中手动ack。