kafka 消费者拉取消息

本文只跟踪消费者拉取消息的流程。对于 java 客户端， kafka 的生产者和消费者复用同一个网络 io 类 NetworkClient。

入口在 KafkaConsumer#pollOnce 中，抽出主要步骤：

// 构造 FetchRequest 请求，将请求对象放入 unsent 集合，等待发送
fetcher.sendFetches();

// 取出 unsent 中的请求，调用 NetworkClient#send，NetworkClinet#poll
client.poll(pollTimeout, nowMs, new PollCondition() {
    @Override
    public boolean shouldBlock() {
        // since a fetch might be completed by the background thread, we need this poll condition
        // to ensure that we do not block unnecessarily in poll()
        return !fetcher.hasCompletedFetches();
    }
});

// 返回数据给用户
return fetcher.fetchedRecords();

Fetcher#sendFetches

public synchronized int sendFetches() {
    // 构造拉取消息请求。从哪个节点，哪个分区，什么位置拉取消息
    Map<Node, FetchSessionHandler.FetchRequestData> fetchRequestMap = prepareFetchRequests();
    for (Map.Entry<Node, FetchSessionHandler.FetchRequestData> entry : fetchRequestMap.entrySet()) {
        final Node fetchTarget = entry.getKey();
        final FetchSessionHandler.FetchRequestData data = entry.getValue();
        //1. 借助 Builder 构造 FetchRequest 对象
        final FetchRequest.Builder request = FetchRequest.Builder
                .forConsumer(this.maxWaitMs, this.minBytes, data.toSend())
                .isolationLevel(isolationLevel)
                .setMaxBytes(this.maxBytes)
                .metadata(data.metadata())
                .toForget(data.toForget());
        if (log.isDebugEnabled()) {
            log.debug("Sending {} {} to broker {}", isolationLevel, data.toString(), fetchTarget);
        }

        
        client.send(fetchTarget, request)
                //4. 给 RequestFutureCompletionHandler.future 添加 RequestFutureListener
                .addListener(new RequestFutureListener<ClientResponse>() {
                    @Override
                    public void onSuccess(ClientResponse resp) {
                        synchronized (Fetcher.this) {
                            FetchResponse response = (FetchResponse) resp.responseBody();
                            FetchSessionHandler handler = sessionHandler(fetchTarget.id());
                            if (handler == null) {
                                log.error("Unable to find FetchSessionHandler for node {}. Ignoring fetch response.",
                                    fetchTarget.id());
                                return;
                            }
                            if (!handler.handleResponse(response)) {
                                return;
                            }

                            Set<TopicPartition> partitions = new HashSet<>(response.responseData().keySet());
                            FetchResponseMetricAggregator metricAggregator = new FetchResponseMetricAggregator(sensors, partitions);

                            for (Map.Entry<TopicPartition, FetchResponse.PartitionData> entry : response.responseData().entrySet()) {
                                TopicPartition partition = entry.getKey();
                                long fetchOffset = data.sessionPartitions().get(partition).fetchOffset;
                                FetchResponse.PartitionData fetchData = entry.getValue();

                                log.debug("Fetch {} at offset {} for partition {} returned fetch data {}",
                                        isolationLevel, fetchOffset, partition, fetchData);
                                // 10. 把数据放入 completedFetches，最终返回给用户
                                completedFetches.add(new CompletedFetch(partition, fetchOffset, fetchData, metricAggregator,
                                        resp.requestHeader().apiVersion()));
                            }

                            sensors.fetchLatency.record(resp.requestLatencyMs());
                        }
                    }

                    @Override
                    public void onFailure(RuntimeException e) {
                        synchronized (Fetcher.this) {
                            FetchSessionHandler handler = sessionHandler(fetchTarget.id());
                            if (handler != null) {
                                handler.handleError(e);
                            }
                        }
                    }
                });
    }
    return fetchRequestMap.size();
}

ConsumerNetworkClient#send

public RequestFuture<ClientResponse> send(Node node, AbstractRequest.Builder<?> requestBuilder) {
    long now = time.milliseconds();
    //2. 使用 RequestFutureCompletionHandler 作为回调函数
    RequestFutureCompletionHandler completionHandler = new RequestFutureCompletionHandler();
    ClientRequest clientRequest = client.newClientRequest(node.idString(), requestBuilder, now, true,
            completionHandler);
    //3. 请求放入 unsent 集合        
    unsent.put(node, clientRequest);

    // wakeup the client in case it is blocking in poll so that we can send the queued request
    client.wakeup();
    return completionHandler.future;
}

ConsumerNetworkClient#poll

// 5. 发送 unsent 中的请求，并没有产生网络 io
trySend(now);

// 真实的网络数据写和读
// 6. 发送请求
// 7. 接收响应
// 8. 触发 RequestFutureCompletionHandler 回调
client.poll(0, now);

// 9. 触发 RequestFutureListener 中的回调
firePendingCompletedRequests();

NetworkClient#handleCompletedReceives

private void handleCompletedReceives(List<ClientResponse> responses, long now) {
    for (NetworkReceive receive : this.selector.completedReceives()) {
        String source = receive.source();
        InFlightRequest req = inFlightRequests.completeNext(source);
        Struct responseStruct = parseStructMaybeUpdateThrottleTimeMetrics(receive.payload(), req.header,
            throttleTimeSensor, now);
        if (log.isTraceEnabled()) {
            log.trace("Completed receive from node {} for {} with correlation id {}, received {}", req.destination,
                req.header.apiKey(), req.header.correlationId(), responseStruct);
        }
        AbstractResponse body = AbstractResponse.parseResponse(req.header.apiKey(), responseStruct);
        if (req.isInternalRequest && body instanceof MetadataResponse)
            metadataUpdater.handleCompletedMetadataResponse(req.header, now, (MetadataResponse) body);
        else if (req.isInternalRequest && body instanceof ApiVersionsResponse)
            handleApiVersionsResponse(responses, req, now, (ApiVersionsResponse) body);
        else
            // 此处给 responses 添加元素
            // return new ClientResponse(header, callback, destination, createdTimeMs, timeMs, false, null, response);
            // 直接把请求的 callback 赋值给响应 
            // 生产者发送消息的 callback，是用户通过参数传入的
            // 消费者拉取消息的 callback，是在 ConsumerNetworkClient#send 指定的，是 RequestFutureCompletionHandler
            responses.add(req.completed(body, now));
    }
}

NetworkClient#completeResponses

private void completeResponses(List<ClientResponse> responses) {
    for (ClientResponse response : responses) {
        try {
            // callback.onComplete(this);
            response.onComplete();
        } catch (Exception e) {
            log.error("Uncaught error in request completion:", e);
        }
    }
}

RequestFutureCompletionHandler#onComplete

public void onComplete(ClientResponse response) {
    this.response = response;
    pendingCompletion.add(this);
}

ConsumerNetworkClient#firePendingCompletedRequests

private void firePendingCompletedRequests() {
    boolean completedRequestsFired = false;
    for (;;) {
        RequestFutureCompletionHandler completionHandler = pendingCompletion.poll();
        if (completionHandler == null)
            break;

        completionHandler.fireCompletion();
        completedRequestsFired = true;
    }

    // wakeup the client in case it is blocking in poll for this future's completion
    if (completedRequestsFired)
        client.wakeup();
}

ConsumerNetworkClient.RequestFutureCompletionHandler#fireCompletion

public void fireCompletion() {
    if (e != null) {
        future.raise(e);
    } else if (response.wasDisconnected()) {
        RequestHeader requestHeader = response.requestHeader();
        int correlation = requestHeader.correlationId();
        log.debug("Cancelled {} request {} with correlation id {} due to node {} being disconnected",
                requestHeader.apiKey(), requestHeader, correlation, response.destination());
        future.raise(DisconnectException.INSTANCE);
    } else if (response.versionMismatch() != null) {
        future.raise(response.versionMismatch());
    } else {
        future.complete(response);
    }
}

RequestFuture#complete

public void complete(T value) {
    try {
        if (value instanceof RuntimeException)
            throw new IllegalArgumentException("The argument to complete can not be an instance of RuntimeException");

        if (!result.compareAndSet(INCOMPLETE_SENTINEL, value))
            throw new IllegalStateException("Invalid attempt to complete a request future which is already complete");
        fireSuccess();
    } finally {
        completedLatch.countDown();
    }
}

private void fireSuccess() {
    T value = value();
    while (true) {
        RequestFutureListener<T> listener = listeners.poll();
        if (listener == null)
            break;
        // 终于调到 RequestFutureListener
        listener.onSuccess(value);
    }
}

如果不考虑心跳线程，consumer 第一次 poll 是不会有数据的，此时请求才发出去，响应还没回来，必须在第二次 poll 时，才能同时处理网络读写事件。

跟完之后，个人觉得调用链还是挺长的。一点感觉，全程只有一个线程，但是每次走的分支都不一样，给人的启发就是，单线程只要不等待，速度也很快。