無法在Kafka的Avro Consumer端解碼自定義對象

Question

我有一個具體的類，正在按字節數組序列化以發送給Kafka主題。 對於序列化，我使用ReflectDatumWriter。 在發送bytes []之前，我在檢查了一些在線教程之后，將模式ID放在具有模式ID的前4個字節中。

我能夠發送消息，但是在Avro控制台使用者中使用它時，得到的響應如下：

./bin/kafka-avro-console-consumer --bootstrap-server 0：9092 --property schema.stry.url = http：// 0：8081 --property print.key = true --topic測試

"1" "\u0000" 
"1" "\u0000" 
"1" "\u0000" 
"1" "\u0000" 
"1" "\u0000"
"1" "\u0000" 
"1" "\u0000" 
"1" "\u0000" 
"1" "\u0000" 
"1" "\u0000"

    MParams ddb = new MParams();
    ddb.setKey("ss");

    for (int i = 0; i < 10; i++) {
        ProducerRecord record = new ProducerRecord<String, byte[]>("Test", "1", build(1, Producer.serialize(ddb)));
        Future resp = kafkaFullAckProducer.send(record);

        System.out.println("Success" + resp.get());
    }
}

public static <T> byte[] serialize(T data) {
    Schema schema = null;
    if (data == null) {
        throw new RuntimeException("Data cannot be null in AvroByteSerializer");
    }
    try {
        schema = ReflectData.get().getSchema(data.getClass());
        ByteArrayOutputStream out = new ByteArrayOutputStream();
        DatumWriter<T> writer = new ReflectDatumWriter<T>(schema);
        writer.write(data, new EncoderFactory().directBinaryEncoder(out, null));
        byte[] bytes = out.toByteArray();
        return bytes;
    } catch (java.io.IOException e) {
        throw new RuntimeException("Error serializing Avro message", e);
    }
}

public static byte[] build(Integer schemaId, byte[] data) {
    ByteArrayOutputStream out = new ByteArrayOutputStream();
    out.write(0);
    try {
        out.write(ByteBuffer.allocate(4).putInt(schemaId).array());
        out.write(data);
        byte[] bytes = out.toByteArray();
        out.close();
        return bytes;
    } catch (IOException e) {
        throw new RuntimeException("Exception in avro record builder , msg :" + e.getMessage());
    }



@Data
public class MParams extends MetricParams{

    // POJO version

    @Nullable
    private String key;


}

@JsonTypeInfo(use = Id.CLASS, include = As.PROPERTY, property = "@c")
@Union(value= {MParams.class})
public abstract class MetricParams {

}

工作的序列化器代碼段

public byte[] serialize(String topic, T record) {
        Schema schema;
        int id;
        try {
            schema = ReflectData.get().getSchema(record.getClass());
            id = client.register(topic + "-value", schema);
        } catch (IOException | RestClientException e) {
            throw new RuntimeException(e);
        }
        return serializeImpl(id, schema, record);
    }

    protected byte[] serializeImpl(int id, Schema schema, T object) throws SerializationException {
        if (object == null) {
            return null;
        }
        try {
            ByteArrayOutputStream out = new ByteArrayOutputStream();
            out.write(0x0);
            out.write(ByteBuffer.allocate(4).putInt(id).array());

            BinaryEncoder encoder = EncoderFactory.get().directBinaryEncoder(out, null);
            DatumWriter<T> writer = new ReflectDatumWriter<T>(schema);
            writer.write(object, encoder);
            encoder.flush();
            byte[] bytes = out.toByteArray();
            out.close();
            return bytes;
        } catch (IOException | RuntimeException e) {
            throw new SerializationException("Error serializing Avro message", e);
        }
    }

解串器：

protected T deserialize(Schema schema, byte[] payload) throws SerializationException {
        // Even if the caller requests schema & version, if the payload is null
        // cannot include it. The caller must handle
        // this case.
        if (payload == null) {
            return null;
        }

        int id = -1;
        try {
            ByteBuffer buffer = getByteBuffer(payload);
            id = buffer.getInt();
            int length = buffer.limit() - 1 - 4;

            int start = buffer.position() + buffer.arrayOffset();
            DatumReader<T> reader = new ReflectDatumReader<T>(schema);
            T res = reader.read(null, new DecoderFactory().binaryDecoder(buffer.array(), start, length, null));
            return res;
        } catch (IOException | RuntimeException e) {
            throw new SerializationException("Error deserializing Avro message for id " + id, e);
        }
    }

    private ByteBuffer getByteBuffer(byte[] payload) {
        ByteBuffer buffer = ByteBuffer.wrap(payload);
        if (buffer.get() != 0x0) {
            throw new SerializationException("Unknown magic byte!");
        }
        return buffer;
    }

Answer 1

對於序列化，我使用ReflectDatumWriter。 在發送bytes []之前，我將模式ID放在前4個具有模式ID的字節中

不清楚為什么要嘗試繞過KafkaAvroSerializer類的默認行為 。 （在您的情況下， Schema.Parser從該示例中刪除Schema.Parser ，並使用Reflect記錄類型，而不是GenericRecord ）

您可以將具體的類作為生產者的第二種類型，並且只要它實現基本的Avro類，就應該正確序列化（這意味着ID正確計算，而不是您創建的一些數字，並轉換為字節），並注冊到注冊表，然后發送給Kafka

最重要的是，架構ID在注冊表中不一定是1，這樣，控制台使用者可能會嘗試錯誤地反序列化消息，從而導致輸出錯誤

換句話說，嘗試

ProducerRecord<String, MParams> record = new ProducerRecord<>(...)