Spring Cloud Stream Kafka>使用Confluent REST Proxy消費Avro消息

Question

我有以下場景：

• 生產者通過Confluent的REST代理（在Confluent的模式注冊表中注冊模式）將Avro編碼的消息發送到Kafka主題，如http://docs.confluent.io/3.0.0/kafka-rest/docs/intro.html#produce中所述-and-消耗-阿夫羅的消息
• 啟用S​​pring Cloud Stream的消息偵聽主題以獲取新消息

我的應用程序如下所示：

@SpringBootApplication
@EnableBinding(Sink.class)
public class MyApplication {
  private static Logger log = LoggerFactory.getLogger(MyApplication.class);

  public static void main(String[] args) {
    SpringApplication.run(MyApplication.class, args);
  }

  @StreamListener(Sink.INPUT)
  public void myMessageSink(MyMessage message) {
    log.info("Received new message: {}", message);
  }
}

而MyMessage是Avro從Avro架構創建的類。

我的application.properties看起來像這樣：

spring.cloud.stream.bindings.input.destination=myTopic
spring.cloud.stream.bindings.input.group=${spring.application.name}
spring.cloud.stream.bindings.input.contentType=application/*+avro

我的問題是，每次收到新消息時，都會拋出以下異常：

org.springframework.messaging.MessagingException: Exception thrown while invoking MyApplication#myMessageSink[1 args]; nested exception is org.apache.avro.AvroRuntimeException: Malformed data. Length is negative: -27
    at org.springframework.cloud.stream.binding.StreamListenerAnnotationBeanPostProcessor$StreamListenerMessageHandler.handleRequestMessage(StreamListenerAnnotationBeanPostProcessor.java:316) ~[spring-cloud-stream-1.1.0.RELEASE.jar:1.1.0.RELEASE]
    at org.springframework.integration.handler.AbstractReplyProducingMessageHandler.handleMessageInternal(AbstractReplyProducingMessageHandler.java:109) ~[spring-integration-core-4.3.2.RELEASE.jar:4.3.2.RELEASE]
    at org.springframework.integration.handler.AbstractMessageHandler.handleMessage(AbstractMessageHandler.java:127) ~[spring-integration-core-4.3.2.RELEASE.jar:4.3.2.RELEASE]
    at org.springframework.integration.dispatcher.AbstractDispatcher.tryOptimizedDispatch(AbstractDispatcher.java:116) ~[spring-integration-core-4.3.2.RELEASE.jar:4.3.2.RELEASE]
    at org.springframework.integration.dispatcher.UnicastingDispatcher.doDispatch(UnicastingDispatcher.java:148) ~[spring-integration-core-4.3.2.RELEASE.jar:4.3.2.RELEASE]
    ...
Caused by: org.apache.avro.AvroRuntimeException: Malformed data. Length is negative: -27
    at org.apache.avro.io.BinaryDecoder.doReadBytes(BinaryDecoder.java:336) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.io.BinaryDecoder.readString(BinaryDecoder.java:263) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.io.ResolvingDecoder.readString(ResolvingDecoder.java:201) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.generic.GenericDatumReader.readString(GenericDatumReader.java:430) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.generic.GenericDatumReader.readString(GenericDatumReader.java:422) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.generic.GenericDatumReader.readMapKey(GenericDatumReader.java:335) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.generic.GenericDatumReader.readMap(GenericDatumReader.java:321) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.generic.GenericDatumReader.readWithoutConversion(GenericDatumReader.java:177) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.specific.SpecificDatumReader.readField(SpecificDatumReader.java:116) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.generic.GenericDatumReader.readRecord(GenericDatumReader.java:230) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.generic.GenericDatumReader.readWithoutConversion(GenericDatumReader.java:174) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.generic.GenericDatumReader.read(GenericDatumReader.java:152) ~[avro-1.8.1.jar:1.8.1]
    at org.apache.avro.generic.GenericDatumReader.read(GenericDatumReader.java:144) ~[avro-1.8.1.jar:1.8.1]
    at org.springframework.cloud.stream.schema.avro.AbstractAvroMessageConverter.convertFromInternal(AbstractAvroMessageConverter.java:91) ~[spring-cloud-stream-schema-1.1.0.RELEASE.jar:1.1.0.RELEASE]
    at org.springframework.messaging.converter.AbstractMessageConverter.fromMessage(AbstractMessageConverter.java:175) ~[spring-messaging-4.3.3.RELEASE.jar:4.3.3.RELEASE]
    at org.springframework.messaging.converter.CompositeMessageConverter.fromMessage(CompositeMessageConverter.java:67) ~[spring-messaging-4.3.3.RELEASE.jar:4.3.3.RELEASE]
    at org.springframework.messaging.handler.annotation.support.PayloadArgumentResolver.resolveArgument(PayloadArgumentResolver.java:117) ~[spring-messaging-4.3.3.RELEASE.jar:4.3.3.RELEASE]
    at org.springframework.messaging.handler.invocation.HandlerMethodArgumentResolverComposite.resolveArgument(HandlerMethodArgumentResolverComposite.java:112) ~[spring-messaging-4.3.3.RELEASE.jar:4.3.3.RELEASE]
    at org.springframework.messaging.handler.invocation.InvocableHandlerMethod.getMethodArgumentValues(InvocableHandlerMethod.java:138) ~[spring-messaging-4.3.3.RELEASE.jar:4.3.3.RELEASE]
    at org.springframework.messaging.handler.invocation.InvocableHandlerMethod.invoke(InvocableHandlerMethod.java:107) ~[spring-messaging-4.3.3.RELEASE.jar:4.3.3.RELEASE]
    at org.springframework.cloud.stream.binding.StreamListenerAnnotationBeanPostProcessor$StreamListenerMessageHandler.handleRequestMessage(StreamListenerAnnotationBeanPostProcessor.java:307) ~[spring-cloud-stream-1.1.0.RELEASE.jar:1.1.0.RELEASE]
    ... 35 common frames omitted

據我所知，問題在於Confluent堆棧包含消息模式的ID作為消息有效負載的一部分，並且客戶端應該在模式ID之后開始讀取實際的Avro消息。 我似乎需要配置Kafka綁定以使用Confluent的KafkaAvroDeserializer，但我無法弄清楚如何實現這一點。

（我可以使用Confluent的avro控制台消費者完全正確地檢索消息，因此它似乎不是Avro編碼的問題）

我也嘗試使用@EnableSchemaRegistry注釋並配置ConfluentSchemaRegistryClient bean，但它在我看來這樣只能控制存儲/檢索模式的位置，而不是實際的反序列化。

這甚至應該以某種方式起作用嗎？

Answer 1

在設置per-binding屬性spring.cloud.stream.kafka.bindings.input.consumer.configuration.value.deserializer以使Confluent's KafkaAvroDeserializer class name是否有效？

Answer 2

回答我自己的問題。 我現在所做的是實現一個MessageConverter，它只是刪除任何消息的前4個字節，然后再將它們傳遞給Avro解碼器。 代碼主要來自spring-cloud-stream的AbstractAvroMessageConverter：

public class ConfluentAvroSchemaMessageConverter extends AvroSchemaMessageConverter {

public ConfluentAvroSchemaMessageConverter() {
    super(new MimeType("application", "avro+confluent"));
}

@Override
protected Object convertFromInternal(Message<?> message, Class<?> targetClass, Object conversionHint) {
    Object result = null;
    try {
        byte[] payload = (byte[]) message.getPayload();

        // byte array to contain the message without the confluent header (first 4 bytes)
        byte[] payloadWithoutConfluentHeader = new byte[payload.length - 4];
        ByteBuffer buf = ByteBuffer.wrap(payload);
        MimeType mimeType = getContentTypeResolver().resolve(message.getHeaders());
        if (mimeType == null) {
            if (conversionHint instanceof MimeType) {
                mimeType = (MimeType) conversionHint;
            }
            else {
                return null;
            }
        }

        // read first 4 bytes and copy the rest to the new byte array
        // see https://groups.google.com/forum/#!topic/confluent-platform/rI1WNPp8DJU
        buf.getInt();
        buf.get(payloadWithoutConfluentHeader);

        Schema writerSchema = resolveWriterSchemaForDeserialization(mimeType);
        Schema readerSchema = resolveReaderSchemaForDeserialization(targetClass);
        DatumReader<Object> reader = getDatumReader((Class<Object>) targetClass, readerSchema, writerSchema);
        Decoder decoder = DecoderFactory.get().binaryDecoder(payloadWithoutConfluentHeader, null);
        result = reader.read(null, decoder);
    }
    catch (IOException e) {
            throw new MessageConversionException(message, "Failed to read payload", e);
    }
    return result;

}

然后，我通過application.properties將傳入的Kafka主題的內容類型設置為application / avro + confluent。

這至少讓我檢索使用Confluent堆棧編碼的消息，但當然它不會以任何方式與架構注冊表交互。