Question

我正在尝试根据ConsumerRecord字段之一上的内容在使用前过滤ConsumerRecord消息。

在应用过滤器之前对消费者记录进行采样（在值中查找GP_ID）：

 ConsumerRecord(topic = jdbc-project, partition = 0, offset = 0, CreateTime = 1551118248440, serialized key size = -1, serialized value size = 69, headers = RecordHeaders(headers = [], isReadOnly = false), key = null, value = {"GP_ID": {"bytes": "@"}, "PROJECT_ID": {"bytes": "\u001E\u008C"}, "START_DATE": 1009843200000, "END_DATE": 1041292800000, "TITLE": "Project- FPH", "STATUS_CODE": "INACTIVE"})

KafkaRecordVO(projectId=7820, gpId=64)

当我在kafkaListenerContainerFactory（）中按如下所示设置recordFilterStrategy时：

@Bean
ConcurrentKafkaListenerContainerFactory<String, GenericRecord> kafkaListenerContainerFactoryProject() {
    ConcurrentKafkaListenerContainerFactory<String, GenericRecord> factory = new ConcurrentKafkaListenerContainerFactory<>();
    factory.setConsumerFactory(consumerFactory());
    factory.setRecordFilterStrategy(new RecordFilterStrategy<String, GenericRecord>() {
        @Override
        public boolean filter(ConsumerRecord<String, GenericRecord> consumerRecord) {
            long gpId= KafkaRecordVO.convertByteBufferToLong(consumerRecord.value().get("GP_ID"));
            if(gpId == 10766L || gpId == 10823L || gpId == 10459L || gpId == 10649L)
                return false;
            else
                return true;
        }
    });
    factory.getContainerProperties().setAckMode(AckMode.MANUAL_IMMEDIATE);
    return factory;
}

KafkaRecordVO.convertByteBufferToLong正在将字节缓冲区的值转换为long值。

这是正确的评估，并返回true / false。

但是，当卡夫卡监听器使用它时，如下所示：

@KafkaListener(id = "project", topics = "jdbc-project", containerFactory = "kafkaListenerContainerFactoryProject")
public void consumeProject(ConsumerRecord<String, GenericRecord> record,Acknowledgment acknowledgment) {
    log.debug(record.toString());
    KafkaRecordVO recordVo = new KafkaRecordVO().projectId(record.value().get("PROJECT_ID"))
                                                .budgetYear(record.value().get("GP_ID"));
    log.debug(recordVo.toString());
}

这将返回删除我过滤的字段值的记录：“ GP_ID”

这些是应用过滤器后生成的示例日志（在值中查找GP_ID）：

ConsumerRecord(topic = jdbc-project, partition = 0, offset = 171275, CreateTime = 1551118279371, serialized key size = -1, serialized value size = 181, headers = RecordHeaders(headers = [], isReadOnly = false), key = null, value = { "GP_ID": {"bytes": ""}, "PROJECT_ID": {"bytes": "\u0005â^"}, "START_DATE": 1470009600000, "END_DATE": 1532995200000, "TITLE": "Project 2016 - 2016", "STATUS_CODE": "INACTIVE"})

KafkaRecordVO(projectId=385630, gpId=0)

我在kafkaListener中通过其他字段在日志中找到此字段：“ GP_ID”：{“ bytes”：“”}。如何不剥离价值？这是怎么了？

更新：当我使用实用程序方法将字节转换为long时，发生了此问题。它正在将位置设置为数组的末尾。这就是为什么它返回空数组：

public static Long convertByteBufferToLong(Object byteBuff) {
    //After adding below line, the issue got resolved
    ByteBuffer buf = ((ByteBuffer) byteBuff).duplicate();
    byte[] arr = new byte[buf.remaining()];
    buf.get(arr);
    BigInteger bi =new BigInteger(1,arr);
    return bi.longValue();
}

Answer 1

您的建议没有任何意义；过滤器适配器具有此代码...

@Override
public void onMessage(ConsumerRecord<K, V> consumerRecord, Acknowledgment acknowledgment, Consumer<?, ?> consumer) {
    if (!filter(consumerRecord)) {
        switch (this.delegateType) {
            case ACKNOWLEDGING_CONSUMER_AWARE:
                this.delegate.onMessage(consumerRecord, acknowledgment, consumer);
                break;
            case ACKNOWLEDGING:
                this.delegate.onMessage(consumerRecord, acknowledgment);
                break;
            case CONSUMER_AWARE:
                this.delegate.onMessage(consumerRecord, consumer);
                break;
            case SIMPLE:
                this.delegate.onMessage(consumerRecord);
        }
    }
    else {
        ackFilteredIfNecessary(acknowledgment);
    }
}

它不对记录进行任何操作。

尝试在调试器中运行以查看是否可以看到正在发生的事情。

Spring Kafka过滤器不过滤消费者记录

1 个答案: