1
votes

I am writing a REST proxy like the confluent rest proxy. It takes a JSON payload, schema subject, and id, and then writes the JSON payload as an Avro object into the stream. When I use kafka-avro-console-consumer to read the message, I am getting "unknown magic byte" errors.

Here is my kafka producer config:

        properties.put("client.id", LocalHostUtils.getLocalHostName(null));

        properties.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, StringSerializer.class);
        properties.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, KafkaAvroSerializer.class);

        properties.put(AbstractKafkaAvroSerDeConfig.AUTO_REGISTER_SCHEMAS, false);
        properties.put(KafkaAvroSerializerConfig.VALUE_SUBJECT_NAME_STRATEGY, RecordNameStrategy.class);

        properties.put("schema.registry.url", configValuesManager.getString("dsp_kafka.schema_registry"));

        if (KafkaUtils.isKafkaEnabled()) {
            this.kafkaProducer = new KafkaProducer<String, Object>(properties);
        }

This is how the REST controller converts the incoming JSON to Avro

        Schema schema = null;
        try {
            schema = schemaRegistryClient.getBySubjectAndID(schemaSubject, schemaId);
        } catch (RestClientException e) {
            throw new IOExceptionWithCause(e);
        }

        log.debug(postContent);
        log.info("Subject/Version {}/{} -> {}", schemaSubject, schemaId, schema);
        Object data = toAvro(schema, postContent);

This is the implementation of the toAvro method:

    Object toAvro(Schema schema, String jsonBody) throws IOException
    {
        DatumReader<Object> reader = new GenericDatumReader<Object>(schema);
        Object object = reader.read(
                null, decoderFactory.jsonDecoder(schema, jsonBody));

        return object;

    }

This object is then passed to the schemaValidatingProducer that I configured with properties given above....

        this.kafkaSchemaValidatingProducer.publish(topic, 0, UUID.randomUUID().toString(), data);

This is the publish method on the kafkaSchemaValidatingProducer

    public void publish(String topic, Integer partition, String key, Object data)
    {
        log.debug("publish topic={} key={} value={}", topic, key, data);

        if (!KafkaUtils.isKafkaEnabled()) {
            log.warn("Kafka is not enabled....");
            return;
        }

        ProducerRecord<String, Object> record = new ProducerRecord<String, Object>(topic, key, data);


        Future<RecordMetadata> metadataFuture = kafkaProducer.send(record, new Callback()
        {
            @Override
            public void onCompletion(RecordMetadata metadata, Exception exception)
            {
                if (exception == null) {
                    log.info(metadata.toString());
                    return;
                }

                log.error("exception", exception);

            }
        });

        kafkaProducer.flush();

    }

this is how I am reading the topic

./bin/kafka-avro-console-consumer --bootstrap-server kafka-broker1:9021 --consumer.config client-ssl.properties --topic schema-validated-topic --property print.key=true --property print.value=true --value-deserializer io.confluent.kafka.serializers.KafkaAvroDeserializer --offset earliest --skip-message-on-error --partition 0 --property schema.registry.url http://schema-regisry

This results in....

[2019-08-26 16:30:36,351] ERROR Error processing message, skipping this message:  (kafka.tools.ConsoleConsumer$:76)
org.apache.kafka.common.errors.SerializationException: Error deserializing Avro message for id -1
Caused by: org.apache.kafka.common.errors.SerializationException: Unknown magic byte!

Any idea why I am getting the "Bad magic number error" ?

1

1 Answers

7
votes

I figured out the problem. It was that I was not specifying the key deserializer in my command.

Here is the command that worked.

./bin/kafka-avro-console-consumer \
--bootstrap-server <bootstrap-server> \
--consumer.config client-ssl.properties \
--property schema.registry.url=<schema-registry-url> \
--topic <name-of-topic> \
--property print.key=true \
--property print.value=true \
--value-deserializer io.confluent.kafka.serializers.KafkaAvroDeserializer \
--key-deserializer org.apache.kafka.common.serialization.StringDeserializer