java - 在 producer.send 期间获取 ProducerFencedException 的原因是什么？

Question

尝试将大约 50K 消息加载到 KAFKA 主题中。在少数运行开始时低于异常但并非一直如此。

org.apache.kafka.common.KafkaException: Cannot execute transactional method because we are in an error state  
at org.apache.kafka.clients.producer.internals.TransactionManager.maybeFailWithError(TransactionManager.java:784) ~[kafka-clients-2.0.0.jar:?]  
at org.apache.kafka.clients.producer.internals.TransactionManager.beginAbort(TransactionManager.java:229) ~[kafka-clients-2.0.0.jar:?]  
at  org.apache.kafka.clients.producer.KafkaProducer.abortTransaction(KafkaProducer.java:679) ~[kafka-clients-2.0.0.jar:?]  
at myPackage.persistUpdatesPostAction(MyCode.java:??) ~[aKafka.jar:?]  
...  
Caused by: org.apache.kafka.common.errors.ProducerFencedException: Producer
attempted an operation with an old epoch. Either there is a newer producer with
the same transactionalId, or the producer's transaction has been expired by the
broker.  

代码块如下:

public void persistUpdatesPostAction(List<Message> messageList ) {
    if ((messageList == null) || (messageList.isEmpty())) {
        return;
    }
    logger.createDebug("Messages in batch(postAction) : "+ messageList.size());
    Producer<String,String> producer = KafkaUtils.getProducer(Thread.currentThread().getName());
    try {
        producer.beginTransaction();
        createKafkaBulkInsert1(producer, messageList, "Topic1");
        createKafkaBulkInsert2(producer, messageList, "Topic2");
        createKafkaBulkInsert3(producer, messageList, "Topic3");
        producer.commitTransaction();
    } catch (Exception e) {
        producer.abortTransaction();
        producer.close();
        KafkaUtils.removeProducer(Thread.currentThread().getName());
    }
}

-----------

static Properties setPropertiesProducer() {
    Properties temp = new Properties();
    temp.put("bootstrap.servers", "localhost:9092");
    temp.put("acks", "all");
    temp.put("retries", 1);
    temp.put("batch.size", 16384);
    temp.put("linger.ms", 5);
    temp.put("buffer.memory", 33554432);
    temp.put("key.serializer",   "org.apache.kafka.common.serialization.StringSerializer");
    temp.put("value.serializer", "org.apache.kafka.common.serialization.StringSerializer");
    return temp;
}

public static Producer<String, String> getProducer(String aThreadId) {
    if ((producerMap.size() == 0) || (producerMap.get(aThreadId) == null)) {
        Properties temp = producerProps;
        temp.put("transactional.id", aThreadId);
        Producer<String, String> producer = new KafkaProducer<String, String>(temp);
        producerMap.put(aThreadId, producer);
        producer.initTransactions();
        return producer;
    }
    return producerMap.get(aThreadId);
}

public static void removeProducer(String aThreadId) {
    logger.createDebug("Removing Thread ID :" + aThreadId);
    if (producerMap.get(aThreadId) == null)
        return;
    producerMap.remove(aThreadId);
}

Answer 1

Caused by: org.apache.kafka.common.errors.ProducerFencedException: Producerattempted an operation with an old epoch. Either there is a newer producer withthe same transactionalId, or the producer's transaction has been expired by thebroker.

此异常消息不是很有帮助。我相信它试图说经纪人不再有客户发送的交易ID的任何记录。这可能是因为:

其他人正在使用相同的事务 ID 并且已经提交了它。根据我的经验，除非您在客户端之间共享事务 ID，否则这种情况不太可能发生。我们使用 UUID.randomUUID() 确保我们的 ID 是唯一的.

交易超时并被经纪人自动化删除。

在我们的例子中，我们经常遇到事务超时，从而产生了这个异常。有 2 个属性控制代理在中止和忘记交易之前记住交易的时间。

transaction.max.timeout.ms -- 一个代理属性，它指定在事务被中止和遗忘之前的最大毫秒数。许多 Kafka 版本的默认值似乎是 900000(15 分钟)。 Documentation from Kafka说:

The maximum allowed timeout for transactions. If a client’s requested transaction time exceeds this, then the broker will return an error in InitProducerIdRequest. This prevents a client from too large of a timeout, which can stall consumers reading from topics included in the transaction.

transaction.timeout.ms -- 一个生产者客户端属性，在创建事务时以毫秒为单位设置超时。许多 Kafka 版本的默认值似乎是 60000(1 分钟)。来自 Kafka 的文档说:

The maximum amount of time in ms that the transaction coordinator will wait for a transaction status update from the producer before proactively aborting the ongoing transaction.

如果 transaction.timeout.ms客户端中设置的属性超过 transaction.max.timeout.ms属性，生产者将立即抛出类似以下异常:

org.apache.kafka.common.KafkaException: Unexpected error in
InitProducerIdResponse The transaction timeout is larger than the maximum value
allowed by the broker (as configured by transaction.max.timeout.ms).