Reactive Kafka:使用事务处理一次 [英] Reactive Kafka : Exactly Once Processing with Transaction

查看:30
本文介绍了Reactive Kafka:使用事务处理一次的处理方法,对大家解决问题具有一定的参考价值,需要的朋友们下面随着小编来一起学习吧!

问题描述

最初通过 api 调用触发

Initially triggered via api call

1. Service A produces m1 to topic1 (non transactional send)

2. Service B consumes topic1 and does some processing
(begin tx)

3. Service B produces m2 to topic2
(commit tx)

4. Service A consumes topic2
(begin tx)

这是我的生产者配置:

final Map<String, Object> props = Maps.newConcurrentMap();

props.put(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG, bootstrapServers);
props.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, StringSerializer.class);
props.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, JsonSerializer.class);


props.put(ProducerConfig.ENABLE_IDEMPOTENCE_CONFIG, "true");
props.put(ProducerConfig.ACKS_CONFIG, "all"); 
props.put(ProducerConfig.TRANSACTIONAL_ID_CONFIG, "producer-tx-1");

这是我的消费者配置:

final Map<String, Object> props = Maps.newHashMap();

props.put(ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG, bootstrapServers);
props.put(ConsumerConfig.GROUP_ID_CONFIG, consumerGroupId);
props.put(ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class);
props.put(ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, JsonDeserializer.class);

props.put(ConsumerConfig.ENABLE_AUTO_COMMIT_CONFIG, "false");
props.put(ConsumerConfig.ISOLATION_LEVEL_CONFIG, "read_committed");

我读了这个示例场景

我尝试遵循但我遇到了一些问题:

And I try to follow but i'm getting some issues:

这是我的生产者代码:

    @Override
public Mono<SenderResult<Void>> buy(Message msg) {
    final ReactiveKafkaProducerTemplate kafkaProducerTemplate = kafkaConfig.getKafkaProducerTemplate();
    return kafkaProducerTemplate.send(mytopic, msg);

}

我的消费者代码:

@Override
public void run(ApplicationArguments arg0) throws Exception {
    final ReactiveKafkaProducerTemplate kafkaProducerTemplate = kafkaConfig.getKafkaProducerTemplate();
    final ReactiveKafkaConsumerTemplate kafkaConsumerTemplate = kafkaConfig.getKafkaConsumerTemplate(mytopic, Message.class);

    final Flux<ConsumerRecord<String, Message>> flux = kafkaConsumerTemplate.receiveExactlyOnce(kafkaProducerTemplate.transactionManager())
            .concatMap(receiverRecordFlux -> receiverRecordFlux );

    flux.subscribe(record -> {
        final Message message = record.value();

    System.out.printf("received message: timestamp=%s key=%d value=%s\n",
                dateFormat.format(new Date(record.timestamp())),
                record.key(),
                message);

     transactionService.processAndSendToNextTopic(message)
                .doOnSuccess(aVoid -> kafkaProducerTemplate.transactionManager().commit())
                .subscribe();

    });
}

在测试以生成和消费消息到单个分区主题的情况下,我总是遇到以下错误.

I'm always getting following error when testing to produce and consume message to a single partitioned topic under happy case.

Invalid transition attempted from state IN_TRANSACTION to state IN_TRANSACTION

有人可以告诉我如何正确开始事务,以及何时提交和关闭事务吗?

Can someone please show me how to correctly begin a transaction, and when to commit and close a transaction?

请注意,如果我不使用事务性 - receive() 而不是 receiveExactlyOnce,这一切都可以正常工作

Note that this all works fine if i don't use transactional - receive() instead of receiveExactlyOnce

推荐答案

您可能需要将 KafkaTransactionManager 同步配置为 SYNCHRONIZATION_ALWAYS.

You may have to configure KafkaTransactionManager synchronization to SYNCHRONIZATION_ALWAYS.

请检查.

另见

Spring Kafka 中的事务同步

https://docs.spring.io/spring-kafka/api/org/springframework/kafka/transaction/KafkaTransactionManager.html

这篇关于Reactive Kafka:使用事务处理一次的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!

查看全文
登录 关闭
扫码关注1秒登录
发送“验证码”获取 | 15天全站免登陆