Kafka源码分析－序列5 －Producer －RecordAccumulator队列分析

在Kafka源码分析－序列2中，我们提到了整个Producer client的架构图，如下所示：

其它几个组件我们在前面都讲过了，今天讲述最后一个组件RecordAccumulator.

Batch发送

在以前的kafka client中，每条消息称为 “Message”，而在Java版client中，称之为”Record”，同时又因为有批量发送累积功能，所以称之为RecordAccumulator.

RecordAccumulator最大的一个特性就是batch消息，扔到队列中的多个消息，可能组成一个RecordBatch，然后由Sender一次性发送出去。

每个TopicPartition一个队列

下面是RecordAccumulator的内部结构，可以看到，每个TopicPartition对应一个消息队列，只有同一个TopicPartition的消息，才可能被batch。

public final class RecordAccumulator {
    private final ConcurrentMap<TopicPartition, Deque<RecordBatch>> batches;

   ...
}

batch的策略

那什么时候，消息会被batch，什么时候不会呢？下面从KafkaProducer的send方法看起：

//KafkaProducer
    public Future<RecordMetadata> send(ProducerRecord<K, V> record, Callback callback) {
        try {
            // first make sure the metadata for the topic is available
            long waitedOnMetadataMs = waitOnMetadata(record.topic(), this.maxBlockTimeMs);

            ...

            RecordAccumulator.RecordAppendResult result = accumulator.append(tp, serializedKey, serializedValue, callback, remainingWaitMs);   //核心函数：把消息放入队列

            if (result.batchIsFull || result.newBatchCreated) {
                log.trace("Waking up the sender since topic {} partition {} is either full or getting a new batch", record.topic(), partition);
                this.sender.wakeup();
            }
            return result.future;

从上面代码可以看到，batch逻辑，都在accumulator.append函数里面：

    public RecordAppendResult append(TopicPartition tp, byte[] key, byte[] value, Callback callback, long maxTimeToBlock) throws InterruptedException {
        appendsInProgress.incrementAndGet();
        try {
            if (closed)
                throw new IllegalStateException("Cannot send after the producer is closed.");
            Deque<RecordBatch> dq = dequeFor(tp);  //找到该topicPartiton对应的消息队列
            synchronized (dq) {
                RecordBatch last = dq.peekLast(); //拿出队列的最后1个元素
                if (last != null) {  
                    FutureRecordMetadata future = last.tryAppend(key, value, callback, time.milliseconds()); //最后一个元素, 即RecordBatch不为空，把该Record加入该RecordBatch
                    if (future != null)
                        return new RecordAppendResult(future, dq.size() > 1 || last.records.isFull(), false);
                }
            }

            int size = Math.max(this.batchSize, Records.LOG_OVERHEAD + Record.recordSize(key, value));
            log.trace("Allocating a new {} byte message buffer for topic {} partition {}", size, tp.topic(), tp.partition());
            ByteBuffer buffer = free.allocate(size, maxTimeToBlock);
            synchronized (dq) {
                // Need to check if producer is closed again after grabbing the dequeue lock.
                if (closed)
                    throw new IllegalStateException("Cannot send after the producer is closed.");
                RecordBatch last = dq.peekLast();
                if (last != null) {
                    FutureRecordMetadata future = last.tryAppend(key, value, callback, time.milliseconds());
                    if (future != null) {
                        // Somebody else found us a batch, return the one we waited for! Hopefully this doesn't happen often...
                        free.deallocate(buffer);
                        return new RecordAppendResult(future, dq.size() > 1 || last.records.isFull(), false);
                    }
                }

                //队列里面没有RecordBatch，建一个新的，然后把Record放进去
                MemoryRecords records = MemoryRecords.emptyRecords(buffer, compression, this.batchSize);
                RecordBatch batch = new RecordBatch(tp, records, time.milliseconds());
                FutureRecordMetadata future = Utils.notNull(batch.tryAppend(key, value, callback, time.milliseconds()));

                dq.addLast(batch);
                incomplete.add(batch);
                return new RecordAppendResult(future, dq.size() > 1 || batch.records.isFull(), true);
            }
        } finally {
            appendsInProgress.decrementAndGet();
        }
    }

    private Deque<RecordBatch> dequeFor(TopicPartition tp) {
        Deque<RecordBatch> d = this.batches.get(tp);
        if (d != null)
            return d;
        d = new ArrayDeque<>();
        Deque<RecordBatch> previous = this.batches.putIfAbsent(tp, d);
        if (previous == null)
            return d;
        else
            return previous;
    }

从上面代码我们可以看出Batch的策略：
1。如果是同步发送，每次去队列取，RecordBatch都会为空。这个时候，消息就不会batch，一个Record形成一个RecordBatch

2。Producer 入队速率 < Sender出队速率 && lingerMs = 0 ，消息也不会被batch

3。Producer 入队速率 > Sender出对速率， 消息会被batch

4。lingerMs > 0，这个时候Sender会等待，直到lingerMs > 0 或者 队列满了，或者超过了一个RecordBatch的最大值，就会发送。这个逻辑在RecordAccumulator的ready函数里面。

    public ReadyCheckResult ready(Cluster cluster, long nowMs) {
        Set<Node> readyNodes = new HashSet<Node>();
        long nextReadyCheckDelayMs = Long.MAX_VALUE;
        boolean unknownLeadersExist = false;

        boolean exhausted = this.free.queued() > 0;
        for (Map.Entry<TopicPartition, Deque<RecordBatch>> entry : this.batches.entrySet()) {
            TopicPartition part = entry.getKey();
            Deque<RecordBatch> deque = entry.getValue();

            Node leader = cluster.leaderFor(part);
            if (leader == null) {
                unknownLeadersExist = true;
            } else if (!readyNodes.contains(leader)) {
                synchronized (deque) {
                    RecordBatch batch = deque.peekFirst();
                    if (batch != null) {
                        boolean backingOff = batch.attempts > 0 && batch.lastAttemptMs + retryBackoffMs > nowMs;
                        long waitedTimeMs = nowMs - batch.lastAttemptMs;
                        long timeToWaitMs = backingOff ? retryBackoffMs : lingerMs;
                        long timeLeftMs = Math.max(timeToWaitMs - waitedTimeMs, 0);
                        boolean full = deque.size() > 1 || batch.records.isFull();
                        boolean expired = waitedTimeMs >= timeToWaitMs;
                        boolean sendable = full || expired || exhausted || closed || flushInProgress();  //关键的一句话
                        if (sendable && !backingOff) {
                            readyNodes.add(leader);
                        } else {

                            nextReadyCheckDelayMs = Math.min(timeLeftMs, nextReadyCheckDelayMs);
                        }
                    }
                }
            }
        }

        return new ReadyCheckResult(readyNodes, nextReadyCheckDelayMs, unknownLeadersExist);
    }

为什么是Deque？

在上面我们看到，消息队列用的是一个“双端队列“，而不是普通的队列。
一端生产，一端消费，用一个普通的队列不就可以吗，为什么要“双端“呢？

这其实是为了处理“发送失败，重试“的问题：当消息发送失败，要重发的时候，需要把消息优先放入队列头部重新发送，这就需要用到双端队列，在头部，而不是尾部加入。

当然，即使如此，该消息发出去的顺序，还是和Producer放进去的顺序不一致了。

评论

1 楼 jiazimo 2017-03-09