RocketMQ源码分析----长轮询

废话

这篇文章主要讲RocketMQ的长轮询,为什么叫长轮询我也不清楚....主要别人这样叫我也这样叫吧,大家明白意思就好。

正文

RcocketMQ消费者的模式是pull模式,也就是会定时向Broker请求消息进行消费。在源码中实现是开启了后台线程不停的去pull(当然会先从队列去PullRequest,队列为空会阻塞),刚研究RocketMQ消费者pull的代码之后不久,有个问题:

  • 如果长时间没有消息,消费者不停的去请求那不就会导致broker负载很高吗

当时没有想太多这个问题,后来才发现的,那么我们先回顾一下Broker在没有消息的时候是怎么处理的,首先当然是先获取消息了(下面代码在PullMessageProcessor.java中)

// ....省略其他代码
final GetMessageResult getMessageResult =
                this.brokerController.getMessageStore().getMessage(requestHeader.getConsumerGroup(), requestHeader.getTopic(),
                        requestHeader.getQueueId(), requestHeader.getQueueOffset(), requestHeader.getMaxMsgNums(), subscriptionData);
// ....省略其他代码

然后看下对Result的判断

switch (getMessageResult.getStatus()) {
                // ....省略其他代码
                case NO_MATCHED_LOGIC_QUEUE:
                case NO_MESSAGE_IN_QUEUE:
                    if (0 != requestHeader.getQueueOffset()) {
                    // ....省略其他代码
                    } else {
                        response.setCode(ResponseCode.PULL_NOT_FOUND);
                    }
                    break;
                case OFFSET_FOUND_NULL:
                    response.setCode(ResponseCode.PULL_NOT_FOUND);
                    break;
                case OFFSET_OVERFLOW_ONE:
                    response.setCode(ResponseCode.PULL_NOT_FOUND);
                    break;

            }

上面是没有找到消息的情况,response是要相应给消费者的,这是为特定的code。
下面看下对response的code的判断

            switch (response.getCode()) {
                case ResponseCode.PULL_NOT_FOUND:
                    if (brokerAllowSuspend && hasSuspendFlag) {
                        // 如果broker开启了长轮询,则将长轮询时间设置为30s(消费者传过来的,默认30s),否则设置为1s
                        long pollingTimeMills = suspendTimeoutMillisLong;
                        if (!this.brokerController.getBrokerConfig().isLongPollingEnable()) {
                            pollingTimeMills = this.brokerController.getBrokerConfig().getShortPollingTimeMills();
                        }

                        String topic = requestHeader.getTopic();
                        long offset = requestHeader.getQueueOffset();
                        int queueId = requestHeader.getQueueId();
                        // 将这次请求的信息包括channel全部封装到PullRequest,并保存到pullRequestTable,即把当前的request hold住
                        PullRequest pullRequest = new PullRequest(request, channel, pollingTimeMills,
                                this.brokerController.getMessageStore().now(), offset, subscriptionData);
                        this.brokerController.getPullRequestHoldService().suspendPullRequest(topic, queueId, pullRequest);
                        response = null;
                        break;
                    }
            }

看到response = null了吧,如果是null,就不会给消费者响应,那么消费者就不能执行相应的回调方法了。
那么这时候又有两个问题:

  • 给或者不给消费者响应对消费者有什么影响?
  • 此时不给响应,那么什么时候会给消费者响应?

先看下第一个问题,第一个问题的答案在MQClientAPIImpl的pullMessageAsync(由于源码里写的异步,所以调用的该异步方法,同步的情况就不一样了)

        this.remotingClient.invokeAsync(addr, request, timeoutMillis, new InvokeCallback() {
            @Override
            public void operationComplete(ResponseFuture responseFuture) {
                RemotingCommand response = responseFuture.getResponseCommand();
                if (response != null) {
                    try {
                        PullResult pullResult = MQClientAPIImpl.this.processPullResponse(response);
                        assert pullResult != null;
                        pullCallback.onSuccess(pullResult);
                    } catch (Exception e) {
                        pullCallback.onException(e);
                    }
                } else {
                    // ....
                }
            }
        });

response==null的在这次的讨论范围,先看下响应了会怎样,会调用回调方法pullcallback,传入Response,这个定义在DefaultMQPushConsumerImpl的pullMessage方法中

// 省略其他代码
PullCallback pullCallback = new PullCallback() {
    @Override
    public void onSuccess(PullResult pullResult) {
        if (pullResult != null) {
            switch (pullResult.getPullStatus()) {
                case FOUND:
                    if (pullResult.getMsgFoundList() == null || pullResult.getMsgFoundList().isEmpty()) {
                        DefaultMQPushConsumerImpl.this.executePullRequestImmediately(pullRequest);
                    } else {
                        if (DefaultMQPushConsumerImpl.this.defaultMQPushConsumer.getPullInterval() > 0) {
                            DefaultMQPushConsumerImpl.this.executePullRequestLater(pullRequest,
                                    DefaultMQPushConsumerImpl.this.defaultMQPushConsumer.getPullInterval());
                        } else {
                            DefaultMQPushConsumerImpl.this.executePullRequestImmediately(pullRequest);
                        }
                    }
                    break;
                case NO_NEW_MSG:
                 DefaultMQPushConsumerImpl.this.executePullRequestImmediately(pullRequest);
                    break;
                case NO_MATCHED_MSG:
                    DefaultMQPushConsumerImpl.this.executePullRequestImmediately(pullRequest);
                    break;
            }
        }
    }

    @Override
    public void onException(Throwable e) {
        DefaultMQPushConsumerImpl.this.executePullRequestLater(pullRequest, PullTimeDelayMillsWhenException);
    }
};

如果收到了响应,到最后会调用executePullRequestImmediately或者executePullRequestLater(底层也是调用executePullRequestImmediately,只不过是过一会采取执行),其作用就是将pullRequest放回队列,嘿嘿,那么第一个问题答案就出来了:

  • 如果是成功,那么当然是没问题,我消费完成,把pullRequest放回队列,后台线程会从队列取出来继续请求下一个消息;
  • 如果是失败,由于Broker没有返回,所以,自然执行不到这里,那么后台线程还是阻塞状态知道返回失败,把队列放回去,才进行下一次尝试

那么接下来就是看下第二个问题的答案是什么了,这个问题嘛,需要先看下《ConsumeQueue介绍和其构建过程》这篇文章,在ConsumeQueue定时构建过程中,有几句代码

    if (BrokerRole.SLAVE != DefaultMessageStore.this.getMessageStoreConfig().getBrokerRole() && DefaultMessageStore.this.brokerConfig.isLongPollingEnable()) {
        DefaultMessageStore.this.messageArrivingListener.arriving(dispatchRequest.getTopic(), dispatchRequest.getQueueId(), dispatchRequest.getConsumeQueueOffset() + 1, dispatchRequest.getTagsCode());
    }

ConsumeQueue在CommitLog有消息写入的时候(即有Producer发送了消息),会进行构建,就会调用上面那行代码,那么这个arriving方法到底干了什么,我们进去看看,核心是调用了PullRequestHoldService的notifyMessageArriving方法

    public void notifyMessageArriving(final String topic, final int queueId, final long maxOffset, final Long tagsCode) {
        String key = this.buildKey(topic, queueId);
        ManyPullRequest mpr = this.pullRequestTable.get(key);
        if (mpr != null) {
            List requestList = mpr.cloneListAndClear();
            if (requestList != null) {
                List replayList = new ArrayList();

                for (PullRequest request : requestList) {
                    long newestOffset = maxOffset;
                    if (newestOffset <= request.getPullFromThisOffset()) {
                        newestOffset = this.brokerController.getMessageStore().getMaxOffsetInQuque(topic, queueId);
                    }

                    Long tmp = tagsCode;
                    if (newestOffset > request.getPullFromThisOffset()) {
                        if (tagsCode == null) {
                            // tmp = getLatestMessageTagsCode(topic, queueId,
                            // maxOffset);
                        }
                        if (this.messageFilter.isMessageMatched(request.getSubscriptionData(), tmp)) {
                            try {
                                this.brokerController.getPullMessageProcessor().excuteRequestWhenWakeup(request.getClientChannel(),
                                        request.getRequestCommand());
                            } catch (RemotingCommandException e) {
                                log.error("", e);
                            }
                            continue;
                        }
                    }
                    // 如果已经超过限制了,那么再次重试,如果还是失败,那么直接返回
                    if (System.currentTimeMillis() >= (request.getSuspendTimestamp() + request.getTimeoutMillis())) {
                        try {
                            this.brokerController.getPullMessageProcessor().excuteRequestWhenWakeup(request.getClientChannel(),
                                    request.getRequestCommand());
                        } catch (RemotingCommandException e) {
                            log.error("", e);
                        }
                        continue;
                    }
                    replayList.add(request);
                }
                if (!replayList.isEmpty()) {
                    mpr.addPullRequest(replayList);
                }
            }
        }
    }

逻辑很简单,先从pullRequestTable中取出对于topic+qid下hold住的PullRequest,如果tag是相符的,那么调用PullMessageProcessor.this.processRequest模拟broker处理消费者的请求消息,并做回响应,这里进行调用如果还是查不到消息,那么不会再hold住Request,直接返回

看到这,第二个问题也得到了答案,整个长轮询的原理也清楚了,总结一下:

  • 消费者会不停的从PullRequest的队列里取request然后想broker请求消息,得到broker的响应后会做相应处理并把PullRequest放回队列以便下一次请求
  • broker在查不到消息的情况下会hold住请求,在ReputMessageService不停构建ConsumeQueue的时候,会拿出hold住的请求进行二次处理

你可能感兴趣的:(RocketMQ源码分析----长轮询)