Spring Boot Retry与Transactional结合时重试未按预期生效问题

问题场景 #

代码结构 #

Kafka监听器组件

@KafkaListener
public void method(){
    service1.method1();
}

服务1

@Transactional
public void method1(){
    service2.method2();
}

服务2

@Transactional
@Retry(Exception.class,3)
public void method2(){
    // 业务逻辑，可能抛出OptimisticLockException
    some logic ,may produce an exception
}

预期与实际偏差 #

• 预期：method2抛出异常时触发Spring Retry重试3次
• 实际：重试未执行，直接抛出Kafka顶层异常：

org.springframework.kafka.KafkaException: Seek to current after exception; nested exception is org.springframework.kafka.listener.ListenerExecutionFailedException: Listener method 'public void iesd.devops.monitoring.consumer.qualitygates.QualityGateTestExecutionEventsConsumer.testExecutionEventsListener(org.apache.kafka.clients.consumer.ConsumerRecord<java.lang.String, iesd.devops.monitoring.entity.qualitygates.TestExecutionEvent>)' threw exception; nested exception is javax.persistence.OptimisticLockException: Row was updated or deleted by another transaction (or unsaved-value mapping was incorrect) : [iesd.devops.monitoring.entity.qualitygates.gates.testgates.TestGatePayload#1]; nested exception is javax.persistence.OptimisticLockException: Row was updated or deleted by another transaction (or unsaved-value mapping was incorrect) : [iesd.devops.monitoring.entity.qualitygates.gates.testgates.TestGatePayload#1]
at org.springframework.kafka.listener.SeekUtils.seekOrRecover(SeekUtils.java:208) ~[spring-kafka-2.9.0.jar:2.9.0]
at org.springframework.kafka.listener.DefaultErrorHandler.handleRemaining(DefaultErrorHandler.java:169) ~[spring-kafka-2.9.0.jar:2.9.0]
at org.springframework.kafka.listener.KafkaMessageListenerContainer$ListenerConsumer.invokeErrorHandler(KafkaMessageListenerContainer.java:2817) ~[spring-kafka-2.9.0.jar:2.9.0]

核心原因 #

1. Retry注解拼写错误：Spring Retry的官方注解是@Retryable，而非代码中的@Retry，这直接导致重试逻辑根本不生效。
2. 事务绑定冲突：service1.method1的事务与service2.method2的事务默认以REQUIRED传播行为绑定，当method2抛出异常，上层事务被标记为回滚状态，无法在同一个事务上下文内触发重试，异常直接冒泡到Kafka监听器。
3. Kafka错误处理器拦截：Kafka监听器的DefaultErrorHandler会捕获所有未处理的异常，执行seek操作并封装为KafkaException抛出，阻断了Retry逻辑的执行流程。

解决步骤 #

1. 修正Retry注解并启用Retry #

• 将@Retry替换为@Retryable，并指定重试参数：

  @Transactional
  @Retryable(value = OptimisticLockException.class, maxAttempts = 3)
  public void method2(){
      some logic ,may produce an exception
  }
  

• 在Spring Boot启动类添加@EnableRetry注解，开启Retry功能：

  @SpringBootApplication
  @EnableRetry
  public class Application {
      public static void main(String[] args) {
          SpringApplication.run(Application.class, args);
      }
  }
  

2. 调整事务传播行为，隔离重试事务 #

修改service2.method2的事务传播行为为REQUIRES_NEW，确保每次重试都开启独立事务，避免与上层事务绑定导致的回滚冲突：

@Transactional(propagation = Propagation.REQUIRES_NEW)
@Retryable(value = OptimisticLockException.class, maxAttempts = 3)
public void method2(){
    some logic ,may produce an exception
}

3. 配置Kafka错误处理器，放行目标异常 #

自定义Kafka的DefaultErrorHandler，让OptimisticLockException不被Kafka拦截，交由Retry逻辑处理：

@Bean
public DefaultErrorHandler kafkaErrorHandler() {
    DefaultErrorHandler errorHandler = new DefaultErrorHandler();
    // 允许OptimisticLockException冒泡到Retry逻辑
    errorHandler.addNotRetryableExceptions(OptimisticLockException.class);
    return errorHandler;
}

4. 可选：将Retry逻辑上移到事务外 #

如果需要更彻底的隔离，可将@Retryable移到Kafka监听器方法或service1.method1上，确保重试在事务启动前触发：

@KafkaListener
@Retryable(value = OptimisticLockException.class, maxAttempts = 3)
public void method(){
    service1.method1();
}

内容的提问来源于stack exchange，提问作者Abhiram Reddy