我尝试在kafka streams（kafka 1.0.1）和spring cloud stream（2.0.0-build-snapshot）的帮助下实现一个简单的事件源服务。我的streamlistener方法只是读取与我的聚合的状态更改相对应的kstream事件，并将它们应用于聚合，并将最新状态保存在本地状态存储（kafka提供的状态存储）中。域事件消息也具有与聚合的uuid（字符串）相同的键。代码如下：

@StreamListener(Channels.EVENTS_INPUT_CHANNEL)
public void listen(KStream<String, DomainEvent> stream) {
    Serde<DomainEvent> domainEventSerde = new JsonSerde<>(DomainEvent.class);
    Serde<Slot> slotSerde = new JsonSerde<>(Slot.class);
    stream
        .groupByKey(Serialized.with(Serdes.String(), domainEventSerde))
        .aggregate(
                Slot::new, 
                (s, domainEvent, slot) -> slot.handle(domainEvent),
                Materialized.<String, Slot, KeyValueStore<Bytes, byte[]>>
                as(Repository.SNAPSHOTS_FOR_SLOTS)
                    .withKeySerde(Serdes.String()).withValueSerde(slotSerde)
        );
}

上面的代码生成一个changelog主题（如预期的那样）：slot service slots changelog。尽管它也创建了一个重新分区主题：插槽服务插槽重新分区。两个主题似乎有完全相同的消息（键和值）。我的理解是，如果没有对流执行键修改操作，则不需要重新分区。我是不是漏了什么？
更新：这可能不再需要了，因为sobychacko已经提供了解释，但是我确实尝试过不使用云流绑定，如下面所示，并且它没有创建重新分区主题：

@Configuration
@EnableKafka
@EnableKafkaStreams
public class KafkaConfiguration {

    @Bean
    KafkaTemplate<String, DomainEvent> kafkaTemplate() {
        return new KafkaTemplate<>(producerFactory());
    }

    @Bean
    ProducerFactory<String,DomainEvent> producerFactory() {
        return new DefaultKafkaProducerFactory<>(config());
    }

    private Map<String, Object> config() {
        Map<String, Object> config = new HashMap<>();
        config.put(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG, "localhost:9092");
        config.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, StringSerializer.class);
        config.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, JsonSerializer.class);
        return config;
    }

    @Bean(name = KafkaStreamsDefaultConfiguration.DEFAULT_STREAMS_CONFIG_BEAN_NAME)
    StreamsConfig streamsConfig() {
        Map<String, Object> config = new HashMap<>();
        config.put(StreamsConfig.BOOTSTRAP_SERVERS_CONFIG, "localhost:9092");
        config.put(StreamsConfig.APPLICATION_ID_CONFIG, "slot-service");
        return new StreamsConfig(config);
    }

    @Bean
    KTable<String, Slot> kTable(KStreamBuilder builder) {
        Serde<DomainEvent> domainEventSerde = new JsonSerde<>(DomainEvent.class);
        Serde<Slot> slotSerde = new JsonSerde<>(Slot.class);

        return
                builder
                .stream(Serdes.String(), domainEventSerde, Repository.SLOT_EVENTS)
                .groupByKey(Serdes.String(), domainEventSerde)
                .aggregate(
                    Slot::new, 
                    (s, domainEvent, slot) -> slot.handle(domainEvent),
                    slotSerde,
                    Repository.SNAPSHOTS_FOR_SLOTS);
    }

    }

此外，生产商如下：

@Autowired
    public Repository(KafkaTemplate<String, DomainEvent> kafkaTemplate, KStreamBuilderFactoryBean kStreamBuilderFactoryBean) {
        this.kafkaTemplate = kafkaTemplate;
        this.kStreamBuilderFactoryBean = kStreamBuilderFactoryBean;
    }

    public void save(Slot slot) {
        List<DomainEvent> newEvents = slot.getDirtyEvents();
        newEvents.forEach(
            domainEvent -> kafkaTemplate.send(SLOT_EVENTS, domainEvent.aggregateUUID().toString(),domainEvent) 
        );
        slot.flushEvents();
    }

更新2：
以下是云流的生产者代码：

public void save(Slot slot) {
        List<DomainEvent> newEvents = slot.getDirtyEvents();
        newEvents.forEach(domainEvent -> channels.eventsOutputChannel().send(MessageBuilder.withPayload(domainEvent)
                .setHeader(KafkaHeaders.MESSAGE_KEY, slot.getUuid().toString()).build()));
        slot.flushEvents();
    }