RocketMq深入分析讲解两种削峰方式
氵奄不死的鱼 人气:0何时需要削峰
当上游调用下游服务速率高于下游服务接口QPS时，那么如果不对调用速率进行控制，那么会发生很多失败请求
通过消息队列的削峰方法有两种
控制消费者消费速率和生产者投放延时消息，本质都是控制消费速度
通过消费者参数控制消费速度
先分析那些参数对控制消费速度有作用
1.PullInterval： 设置消费端，拉取mq消息的间隔时间。
注意：该时间算起时间是rocketMq消费者从broker消息后算起。经过PullInterval再次向broker拉去消息
源码分析:
首先需要了解rocketMq的消息拉去过程
拉去消息的类
PullMessageService
public class PullMessageService extends ServiceThread {
private final InternalLogger log = ClientLogger.getLog();
private final LinkedBlockingQueue<PullRequest> pullRequestQueue = new LinkedBlockingQueue<PullRequest>();
private final MQClientInstance mQClientFactory;
private final ScheduledExecutorService scheduledExecutorService = Executors
.newSingleThreadScheduledExecutor(new ThreadFactory() {
@Override
public Thread newThread(Runnable r) {
return new Thread(r, "PullMessageServiceScheduledThread");
}
});
public PullMessageService(MQClientInstance mQClientFactory) {
this.mQClientFactory = mQClientFactory;
}
public void executePullRequestLater(final PullRequest pullRequest, final long timeDelay) {
if (!isStopped()) {
this.scheduledExecutorService.schedule(new Runnable() {
@Override
public void run() {
PullMessageService.this.executePullRequestImmediately(pullRequest);
}
}, timeDelay, TimeUnit.MILLISECONDS);
} else {
log.warn("PullMessageServiceScheduledThread has shutdown");
}
}
public void executePullRequestImmediately(final PullRequest pullRequest) {
try {
this.pullRequestQueue.put(pullRequest);
} catch (InterruptedException e) {
log.error("executePullRequestImmediately pullRequestQueue.put", e);
}
}
public void executeTaskLater(final Runnable r, final long timeDelay) {
if (!isStopped()) {
this.scheduledExecutorService.schedule(r, timeDelay, TimeUnit.MILLISECONDS);
} else {
log.warn("PullMessageServiceScheduledThread has shutdown");
}
}
public ScheduledExecutorService getScheduledExecutorService() {
return scheduledExecutorService;
}
private void pullMessage(final PullRequest pullRequest) {
final MQConsumerInner consumer = this.mQClientFactory.selectConsumer(pullRequest.getConsumerGroup());
if (consumer != null) {
DefaultMQPushConsumerImpl impl = (DefaultMQPushConsumerImpl) consumer;
impl.pullMessage(pullRequest);
} else {
log.warn("No matched consumer for the PullRequest {}, drop it", pullRequest);
}
}
@Override
public void run() {
log.info(this.getServiceName() + " service started");
while (!this.isStopped()) {
try {
PullRequest pullRequest = this.pullRequestQueue.take();
this.pullMessage(pullRequest);
} catch (InterruptedException ignored) {
} catch (Exception e) {
log.error("Pull Message Service Run Method exception", e);
}
}
log.info(this.getServiceName() + " service end");
}
@Override
public void shutdown(boolean interrupt) {
super.shutdown(interrupt);
ThreadUtils.shutdownGracefully(this.scheduledExecutorService, 1000, TimeUnit.MILLISECONDS);
}
@Override
public String getServiceName() {
return PullMessageService.class.getSimpleName();
}
}继承自ServiceThread，这是一个单线程执行的service，不断获取阻塞队列中的pullRequest，进行消息拉取。
executePullRequestLater会延时将pullrequest放入到pullRequestQueue，达到延时拉去的目的。
那么PullInterval参数就是根据这个功能发挥的作用，在消费者拉去消息成功的回调
PullCallback pullCallback = new PullCallback() {
@Override
public void onSuccess(PullResult pullResult) {
if (pullResult != null) {
pullResult = DefaultMQPushConsumerImpl.this.pullAPIWrapper.processPullResult(pullRequest.getMessageQueue(), pullResult,
subscriptionData);
switch (pullResult.getPullStatus()) {
case FOUND:
long prevRequestOffset = pullRequest.getNextOffset();
pullRequest.setNextOffset(pullResult.getNextBeginOffset());
long pullRT = System.currentTimeMillis() - beginTimestamp;
DefaultMQPushConsumerImpl.this.getConsumerStatsManager().incPullRT(pullRequest.getConsumerGroup(),
pullRequest.getMessageQueue().getTopic(), pullRT);
long firstMsgOffset = Long.MAX_VALUE;
if (pullResult.getMsgFoundList() == null || pullResult.getMsgFoundList().isEmpty()) {
DefaultMQPushConsumerImpl.this.executePullRequestImmediately(pullRequest);
} else {
firstMsgOffset = pullResult.getMsgFoundList().get(0).getQueueOffset();
DefaultMQPushConsumerImpl.this.getConsumerStatsManager().incPullTPS(pullRequest.getConsumerGroup(),
pullRequest.getMessageQueue().getTopic(), pullResult.getMsgFoundList().size());
boolean dispatchToConsume = processQueue.putMessage(pullResult.getMsgFoundList());
DefaultMQPushConsumerImpl.this.consumeMessageService.submitConsumeRequest(
pullResult.getMsgFoundList(),
processQueue,
pullRequest.getMessageQueue(),
dispatchToConsume);
if (DefaultMQPushConsumerImpl.this.defaultMQPushConsumer.getPullInterval() > 0) {
DefaultMQPushConsumerImpl.this.executePullRequestLater(pullRequest,
DefaultMQPushConsumerImpl.this.defaultMQPushConsumer.getPullInterval());
} else {
DefaultMQPushConsumerImpl.this.executePullRequestImmediately(pullRequest);
}
}
if (pullResult.getNextBeginOffset() < prevRequestOffset
|| firstMsgOffset < prevRequestOffset) {
log.warn(
"[BUG] pull message result maybe data wrong, nextBeginOffset: {} firstMsgOffset: {} prevRequestOffset: {}",
pullResult.getNextBeginOffset(),
firstMsgOffset,
prevRequestOffset);
}
break;
case NO_NEW_MSG:
pullRequest.setNextOffset(pullResult.getNextBeginOffset());
DefaultMQPushConsumerImpl.this.correctTagsOffset(pullRequest);
DefaultMQPushConsumerImpl.this.executePullRequestImmediately(pullRequest);
break;
case NO_MATCHED_MSG:
pullRequest.setNextOffset(pullResult.getNextBeginOffset());
DefaultMQPushConsumerImpl.this.correctTagsOffset(pullRequest);
DefaultMQPushConsumerImpl.this.executePullRequestImmediately(pullRequest);
break;
case OFFSET_ILLEGAL:
log.warn("the pull request offset illegal, {} {}",
pullRequest.toString(), pullResult.toString());
pullRequest.setNextOffset(pullResult.getNextBeginOffset());
pullRequest.getProcessQueue().setDropped(true);
DefaultMQPushConsumerImpl.this.executeTaskLater(new Runnable() {
@Override
public void run() {
try {
DefaultMQPushConsumerImpl.this.offsetStore.updateOffset(pullRequest.getMessageQueue(),
pullRequest.getNextOffset(), false);
DefaultMQPushConsumerImpl.this.offsetStore.persist(pullRequest.getMessageQueue());
DefaultMQPushConsumerImpl.this.rebalanceImpl.removeProcessQueue(pullRequest.getMessageQueue());
log.warn("fix the pull request offset, {}", pullRequest);
} catch (Throwable e) {
log.error("executeTaskLater Exception", e);
}
}
}, 10000);
break;
default:
break;
}
}
}
@Override
public void onException(Throwable e) {
if (!pullRequest.getMessageQueue().getTopic().startsWith(MixAll.RETRY_GROUP_TOPIC_PREFIX)) {
log.warn("execute the pull request exception", e);
}
DefaultMQPushConsumerImpl.this.executePullRequestLater(pullRequest, PULL_TIME_DELAY_MILLS_WHEN_EXCEPTION);
}
};在 case found的情况下,也就是拉取到消息的q情况，在PullInterval>0的情况下，会延时投递到pullRequestQueue中，实现拉取消息的间隔
if (DefaultMQPushConsumerImpl.this.defaultMQPushConsumer.getPullInterval() > 0) {
DefaultMQPushConsumerImpl.this.executePullRequestLater(pullRequest,
DefaultMQPushConsumerImpl.this.defaultMQPushConsumer.getPullInterval());
} else {
DefaultMQPushConsumerImpl.this.executePullRequestImmediately(pullRequest);
}
2.PullBatchSize: 设置每次pull消息的数量，该参数设置是针对逻辑消息队列，并不是每次pull消息拉到的总消息数
消费端分配了两个消费队列来监听。那么PullBatchSize 设置为32，那么该消费端每次pull到 64个消息。
消费端每次pull到消息总数=PullBatchSize*监听队列数
源码分析
消费者拉取消息时
org.apache.rocketmq.client.impl.consumer.DefaultMQPushConsumerImpl#pullMessage中
会执行
this.pullAPIWrapper.pullKernelImpl(
pullRequest.getMessageQueue(),
subExpression,
subscriptionData.getExpressionType(),
subscriptionData.getSubVersion(),
pullRequest.getNextOffset(),
this.defaultMQPushConsumer.getPullBatchSize(),
sysFlag,
commitOffsetValue,
BROKER_SUSPEND_MAX_TIME_MILLIS,
CONSUMER_TIMEOUT_MILLIS_WHEN_SUSPEND,
CommunicationMode.ASYNC,
pullCallback
);
其中 this.defaultMQPushConsumer.getPullBatchSize(),就是配置的PullBatchSize，代表的是每次从broker的一个队列上拉取的最大消息数。
3.ThreadMin和ThreadMax： 消费端消费pull到的消息需要的线程数量。
源码分析：
还是在消费者拉取消息成功时
boolean dispatchToConsume = processQueue.putMessage(pullResult.getMsgFoundList());
DefaultMQPushConsumerImpl.this.consumeMessageService.submitConsumeRequest(
pullResult.getMsgFoundList(),
processQueue,
pullRequest.getMessageQueue(),
dispatchToConsume);
通过consumeMessageService执行
默认情况下是并发消费
org.apache.rocketmq.client.impl.consumer.ConsumeMessageConcurrentlyService#submitConsumeRequest
@Override
public void submitConsumeRequest(
final List<MessageExt> msgs,
final ProcessQueue processQueue,
final MessageQueue messageQueue,
final boolean dispatchToConsume) {
final int consumeBatchSize = this.defaultMQPushConsumer.getConsumeMessageBatchMaxSize();
if (msgs.size() <= consumeBatchSize) {
ConsumeRequest consumeRequest = new ConsumeRequest(msgs, processQueue, messageQueue);
try {
this.consumeExecutor.submit(consumeRequest);
} catch (RejectedExecutionException e) {
this.submitConsumeRequestLater(consumeRequest);
}
} else {
for (int total = 0; total < msgs.size(); ) {
List<MessageExt> msgThis = new ArrayList<MessageExt>(consumeBatchSize);
for (int i = 0; i < consumeBatchSize; i++, total++) {
if (total < msgs.size()) {
msgThis.add(msgs.get(total));
} else {
break;
}
}
ConsumeRequest consumeRequest = new ConsumeRequest(msgThis, processQueue, messageQueue);
try {
this.consumeExecutor.submit(consumeRequest);
} catch (RejectedExecutionException e) {
for (; total < msgs.size(); total++) {
msgThis.add(msgs.get(total));
}
this.submitConsumeRequestLater(consumeRequest);
}
}
}
}
其中consumeExecutor初始化
this.consumeExecutor = new ThreadPoolExecutor(
this.defaultMQPushConsumer.getConsumeThreadMin(),
this.defaultMQPushConsumer.getConsumeThreadMax(),
1000 * 60,
TimeUnit.MILLISECONDS,
this.consumeRequestQueue,
new ThreadFactoryImpl("ConsumeMessageThread_"));
对象线程池最大和核心线程数。对于顺序消费ConsumeMessageOrderlyService也会使用最大和最小线程数这两个参数，只是消费时会锁定队列。
以上三种情况：是针对参数配置，来调整消费速度。
除了这三种情况外还有两种服务部署情况，可以调整消费速度：
4.rocketMq 逻辑消费队列配置数量 有消费端每次pull到消息总数=PullBatchSize*监听队列数
可知rocketMq 逻辑消费队列配置数量即上图中的 queue1 ，queue2，配置数量越多每次pull到的消息总数也就越多。如果下边配置读队列数量：修改tocpic的逻辑队列数量
5.消费端节点部署数量 ：
部署数量无论一个节点监听所有队列，还是多个节点按照分配策略分配监听队列数量，理论上每秒pull到的数量都一样的，但是多节点消费端消费线程数量要比单节点消费线程数量多，也就是多节点消费速度大于单节点。
消费延时控流
针对消息订阅者的消费延时流控的基本原理是，每次消费时在客户端增加一个延时来控制消费速度，此时理论上消费并发最快速度为：
单节点部署：
ConsumInterval ：延时时间单位毫秒
ConcurrentThreadNumber：消费端线程数量
MaxRate ：理论每秒处理数量
MaxRate = 1 / ConsumInterval * ConcurrentThreadNumber
如果消息并发消费线程（ConcurrentThreadNumber）为 20，延时（ConsumInterval）为 100 ms，代入上述公式可得
如果消息并发消费线程（ConcurrentThreadNumber）为 20，延时（ConsumInterval）为 100 ms，代入上述公式可得
200 = 1 / 0.1 * 20
由上可知，理论上可以将并发消费控制在 200 以下
如果是多个节点部署如两个节点，理论消费速度最高为每秒处理400个消息。
如下延时流控代码：
/**
* 测试mq 并发 接受
*/
@Component
@RocketMQMessageListener(topic = ConstantTopic.WRITING_LIKE_TOPIC,selectorExpression = ConstantTopic.WRITING_LIKE_ADD_TAG, consumerGroup = "writing_like_topic_add_group")
class ConsumerLikeSave implements RocketMQListener<LikeWritingParams>, RocketMQPushConsumerLifecycleListener{
@SneakyThrows
@Override
public void onMessage(LikeWritingParams params) {
System.out.println("睡上0.1秒");
Thread.sleep(100);
long begin = System.currentTimeMillis();
System.out.println("mq消费速度"+Thread.currentThread().getName()+" "+DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss.SSS").format(LocalDateTime.now()));
//writingLikeService.saveLike2Db(params.getUserId(),params.getWritingId());
long end = System.currentTimeMillis();
// System.out.println("消费：: " +Thread.currentThread().getName()+ "毫秒："+(end - begin));
}
@Override
public void prepareStart(DefaultMQPushConsumer defaultMQPushConsumer) {
defaultMQPushConsumer.setConsumeThreadMin(20); //消费端拉去到消息以后分配线索去消费
defaultMQPushConsumer.setConsumeThreadMax(50);//最大消费线程，一般情况下，默认队列没有塞满，是不会启用新的线程的
defaultMQPushConsumer.setPullInterval(0);//消费端多久一次去rocketMq 拉去消息
defaultMQPushConsumer.setPullBatchSize(32); //消费端每个队列一次拉去多少个消息，若该消费端分赔了N个监控队列，那么消费端每次去rocketMq拉去消息说为N*1
defaultMQPushConsumer.setConsumeFromWhere(ConsumeFromWhere.CONSUME_FROM_TIMESTAMP);
defaultMQPushConsumer.setConsumeTimestamp(UtilAll.timeMillisToHumanString3(System.currentTimeMillis()));
defaultMQPushConsumer.setConsumeMessageBatchMaxSize(2);
}
}注释：如上消费端，单节点每秒处理速度也就是最高200个消息，实际上要小于200，业务代码执行也是需要时间。
但是要注意实际操作中并发流控实际是默认存在的，
spring boot 消费端默认配置
this.consumeThreadMin = 20;
this.consumeThreadMax = 20;
this.pullInterval = 0L;
this.pullBatchSize = 32;
若业务逻辑执行需要20ms，那么单节点处理速度就是：1/0.02*20=1000
这里默认拉去的速度1s内远大于1000
注意： 这里虽然pullInterval 等于0 当时受限于每次拉去64个，处理完也是需要一端时间才能回复ack，才能再次拉取，所以消费速度应该小于1000
所以并发流控要消费速度大于消费延时流控 ，那么消费延时流控才有意义
使用rokcetMq支持的延时消息也可以实现消息的延时消费，通过对delayLevel对应的时间进行配置为我们的需求。为不同的消息设置不同delayLevel，达到延时消费的目的。
总结
rocketMq 肖锋流控两种方式：
并发流控：就是根据业务流控速率要求，来调整topic 消费队列数量(read queue),消费端部署节点，消费端拉去间隔时间，消费端消费线程数量等，来达到要求的速率内
延时消费流控：就是在消费端延时消费消息(sleep),具体延时多少要根据业务要求速率，和消费端线程数量，和节点部署数量来控制
加载全部内容