Kafka生产者发送流程源码解析

1.说明

本文章简单流程追踪一下Kafka从客户端数据发送到Server的流程。
看完本文，你将会大致了解客户端数据发送的过程。

2.发送数据的例子

public class CustomProducer {

    public static void main(String[] args) throws Exception{
        Properties properties = new Properties();
        properties.put(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG, "xxxx:9092");
        properties.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, StringSerializer.class);
        properties.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, StringSerializer.class);

        KafkaProducer<String, String> kafkaProducer = new KafkaProducer<>(properties);
        for (int i = 0; i < 5; i++) {
            kafkaProducer.send(new ProducerRecord<>("test", "hello! --->" + i));
        }
        kafkaProducer.close();
    }
}

以上是我们使用kafka最简单发送数据的例子，我们重点将围绕send()方法进行探究。

3.KafkaProducer

KafkaProducer kafkaProducer = new KafkaProducer<>(properties);
发送数据之间，我们是先实例化了KafkaProducer类，我们大致查看构造器的源码：

public KafkaProducer(Properties properties) {
    this(propsToMap(properties), null, null, null, null, null, Time.SYSTEM);
}

KafkaProducer(Map<String, Object> configs,
               ....) {
     ProducerConfig config = new ProducerConfig(ProducerConfig.addSerializerToConfig(configs, keySerializer,valueSerializer));
     try {
     		// 分区器配置
         this.partitioner = config.getConfiguredInstance(ProducerConfig.PARTITIONER_CLASS_CONFIG, Partitioner.class);
         // 重试的时间间隔，默认100ms
         long retryBackoffMs = config.getLong(ProducerConfig.RETRY_BACKOFF_MS_CONFIG);
       	// 序列化器
         this.keySerializer = config.getConfiguredInstance(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG,
                                                                      Serializer.class);
         this.valueSerializer = config.getConfiguredInstance(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG,Serializer.class);
         // 拦截器
         List<ProducerInterceptor<K, V>> interceptorList = (List) configWithClientId.getConfiguredInstances(
                 ProducerConfig.INTERCEPTOR_CLASSES_CONFIG, ProducerInterceptor.class);
         // 单条消息的最大值，默认1M        
         this.maxRequestSize = config.getInt(ProducerConfig.MAX_REQUEST_SIZE_CONFIG);
         // 缓存的大小,默认32M
         this.totalMemorySize = config.getLong(ProducerConfig.BUFFER_MEMORY_CONFIG);
         // 压缩配置
         this.compressionType = CompressionType.forName(config.getString(ProducerConfig.COMPRESSION_TYPE_CONFIG));
				// 重要，发送数据的累加器
         this.accumulator = new RecordAccumulator(logContext,
                 config.getInt(ProducerConfig.BATCH_SIZE_CONFIG), // 批次大小16K
                 this.compressionType,// 没指定就是none
                 lingerMs(config), // linger.ms 默认0
                 retryBackoffMs, // 重试间隔时间,默认100ms
                 deliveryTimeoutMs, // 默认2分钟
                 metrics,
                 PRODUCER_METRIC_GROUP_NAME,
                 new BufferPool(this.totalMemorySize);
			  // 重要，真正发送数据的任务            
         this.sender = newSender(logContext, kafkaClient, this.metadata);
         String ioThreadName = NETWORK_THREAD_PREFIX + " | " + clientId;
         this.ioThread = new KafkaThread(ioThreadName, this.sender, true);
         // 生产者线程启动
         this.ioThread.start();
         log.debug("Kafka producer started");
     } catch (Throwable t) {
         close(Duration.ofMillis(0), true);
         throw new KafkaException("Failed to construct kafka producer", t);
     }
 }

构造器里面，初始化了一些默认的参数设置，比较重要的是启动了Send线程。

Sender newSender(LogContext logContext, KafkaClient kafkaClient, ProducerMetadata metadata) {
		// 缓存的发送请求默认是5
    int maxInflightRequests = configureInflightRequests(producerConfig, transactionManager != null);
    // 请求超时时间,默认30s
    int requestTimeoutMs = producerConfig.getInt(ProducerConfig.REQUEST_TIMEOUT_MS_CONFIG);
    // 发送数据的客户端
    KafkaClient client = kafkaClient != null ? kafkaClient : new NetworkClient(
            new Selector(producerConfig.getLong(ProducerConfig.CONNECTIONS_MAX_IDLE_MS_CONFIG),
                    this.metrics, time, "producer", channelBuilder, logContext),
            metadata,
            clientId,
            maxInflightRequests,
            producerConfig.getLong(ProducerConfig.RECONNECT_BACKOFF_MS_CONFIG),
            producerConfig.getLong(ProducerConfig.RECONNECT_BACKOFF_MAX_MS_CONFIG),
            producerConfig.getInt(ProducerConfig.SEND_BUFFER_CONFIG),
            producerConfig.getInt(ProducerConfig.RECEIVE_BUFFER_CONFIG),
            requestTimeoutMs,
            time,
            true,
            apiVersions,
            throttleTimeSensor,
            logContext);
    return new Sender(logContext,
            client, // 客户端
            metadata,
            this.accumulator, // 累加器
            maxInflightRequests == 1,
            producerConfig.getInt(ProducerConfig.MAX_REQUEST_SIZE_CONFIG),
            acks,
            retries,
            metricsRegistry.senderMetrics,
            time,
            requestTimeoutMs,
            producerConfig.getLong(ProducerConfig.RETRY_BACKOFF_MS_CONFIG),
            this.transactionManager,
            apiVersions);
}

创建Send线程时，传入了各种配置参数，这些参数都是可以根据实际情况进行调优设置的。

4.send()流程

public Future<RecordMetadata> send(ProducerRecord<K, V> record){
    return send(record, null);
}

public Future<RecordMetadata> send(ProducerRecord<K, V> record, Callback callback) {
    // 第一步会先执行我们配置的拦截器
    ProducerRecord<K, V> interceptedRecord = this.interceptors.onSend(record);
    return doSend(interceptedRecord, callback);
}

doSend(interceptedRecord, callback):

private Future<RecordMetadata> doSend(ProducerRecord<K, V> record, Callback callback) {
 		 TopicPartition tp = null;
     try {
         throwIfProducerClosed();
         // first make sure the metadata for the topic is available
         long nowMs = time.milliseconds();
         ClusterAndWaitTime clusterAndWaitTime;
         try {
             // 从kafka拉取元数据, maxBlockTimeMs 最多等待的时间
             clusterAndWaitTime = waitOnMetadata(record.topic(), record.partition(), nowMs, maxBlockTimeMs);
         } catch (KafkaException e) {
             if (metadata.isClosed())
                 throw new KafkaException("Producer closed while send in progress", e);
             throw e;
         }
         nowMs += clusterAndWaitTime.waitedOnMetadataMs;
         // 剩余时间 = 最多能等待时间 - 用了多少时间；
         long remainingWaitMs = Math.max(0, maxBlockTimeMs - clusterAndWaitTime.waitedOnMetadataMs);
         Cluster cluster = clusterAndWaitTime.cluster;
         // 序列化操作
         byte[] serializedKey;
         serializedKey = keySerializer.serialize(record.topic(), record.headers(), record.key());
         byte[] serializedValue;
         serializedValue = valueSerializer.serialize(record.topic(), record.headers(), record.value());
         // 分区器执行，分区计算
         int partition = partition(record, serializedKey, serializedValue, cluster);
         // 要发送的topic的分区
         tp = new TopicPartition(record.topic(), partition);
         Header[] headers = record.headers().toArray();
         int serializedSize = AbstractRecords.estimateSizeInBytesUpperBound(apiVersions.maxUsableProduceMagic(),
                 compressionType, serializedKey, serializedValue, headers);
         // 检查发送前数据的大小 serializedSize = key + value + header
         ensureValidRecordSize(serializedSize);         
         // 往累加器里面追加要发送的数据
         RecordAccumulator.RecordAppendResult result = accumulator.append(tp, timestamp, serializedKey,
                 serializedValue, headers, interceptCallback, remainingWaitMs, true, nowMs);

         // 批次满了  或者 创建了一个新的批次，唤醒sender发送线程
         if (result.batchIsFull || result.newBatchCreated) {
             this.sender.wakeup();
         }
         return result.future;
     } 

}

send()方法的大致流程就是:
拦截器执行 —> 序列化器 —> 分区器 —> 添加到 RecordAccumulator 里面。

我们在看一下 accumulator.append()方法：

ByteBuffer buffer = null;
try {
    // 根据分区，来创建一个双端队列
    // 每个分区都对应一个队列
    Deque<ProducerBatch> dq = getOrCreateDeque(tp);
    byte maxUsableMagic = apiVersions.maxUsableProduceMagic();
    // batchSize 默认16kb,实际数据的大小
    int size = Math.max(this.batchSize, AbstractRecords.estimateSizeInBytesUpperBound(maxUsableMagic, compression, key, value, headers));
    // 给要发送的数据分配空间
    buffer = free.allocate(size, maxTimeToBlock);
    nowMs = time.milliseconds();
    synchronized (dq) {
        if (closed)
            throw new KafkaException("Producer closed while send in progress");
        // 尝试向队列里面添加数据（有内存，但是没有批次对象）
        RecordAppendResult appendResult = tryAppend(timestamp, key, value, headers, callback, dq, nowMs);
        MemoryRecordsBuilder recordsBuilder = recordsBuilder(buffer, maxUsableMagic);
        // 根据内存大小封装批次（有内存、有批次对象）
        ProducerBatch batch = new ProducerBatch(tp, recordsBuilder, nowMs);
        FutureRecordMetadata future = Objects.requireNonNull(batch.tryAppend(timestamp, key, value, headers,
        callback, nowMs));
        // 把创建的批次加入到队列尾部
        dq.addLast(batch);
        incomplete.add(batch);
        buffer = null;
        return new RecordAppendResult(future, dq.size() > 1 || batch.isFull(), true, false);
    }
} finally {
    if (buffer != null)
        free.deallocate(buffer);
    appendsInProgress.decrementAndGet();
}

RecordAccumulator 类里面的结构：

//            Map
private final ConcurrentMap<TopicPartition, Deque<ProducerBatch>> batches;

以上数据已经被暂存到了累加器里面，我们下面来看看数据如何发送给Server的。
先前开启了一个Sender线程，我们来分析一下这个线程任务：

5.Sender 流程

Sender 实现了 Runnable，直接看run()方法流程：

@Override
public void run() {
	  log.debug("Starting Kafka producer I/O thread.");
    while (running) {
        try {
            runOnce();
        } catch (Exception e) {
            log.error("Uncaught error in kafka producer I/O thread: ", e);
        }
    }
}

void runOnce() {
    long currentTimeMs = time.milliseconds();
    // 准备发送数据给server
    long pollTimeout = sendProducerData(currentTimeMs);
    // 等待发送响应
    client.poll(pollTimeout, currentTimeMs);
}

接下来看一下数据是怎么发送的:
sendProducerData():

private long sendProducerData(long now) {
    // 获取元数据
    Cluster cluster = metadata.fetch();
    // 检测数据是否准备好
    RecordAccumulator.ReadyCheckResult result = this.accumulator.ready(cluster, now);
    if (!result.unknownLeaderTopics.isEmpty()) {
        for (String topic : result.unknownLeaderTopics)
            this.metadata.add(topic, now);
        this.metadata.requestUpdate();
    }
    // create produce requests
    // 将发往同一个节点的数据，放在一个批次里面
    Map<Integer, List<ProducerBatch>> batches = this.accumulator.drain(cluster, result.readyNodes, this.maxRequestSize, now);
    addToInflightBatches(batches);
    // 发送数据
    sendProduceRequests(batches, now);
    return pollTimeout;
}

sendProduceRequests(batches, now):

private void sendProduceRequest(long now, int destination, short acks, int timeout, List<ProducerBatch> batches) {
ClientRequest clientRequest = client.newClientRequest(nodeId, requestBuilder, now, acks != 0,
                requestTimeoutMs, callback);
client.send(clientRequest, now);
}

client 为 NetworkClient#send：

doSend(clientRequest, isInternalRequest, now, builder.build(version));
  |
  |
selector.send(send);

数据发送完成。

6.原理图

完结。