Spring Cloud Eureka源码分析之心跳续约及自我保

2022-01-07  本文已影响0人  跟着Mic学架构
Eureka

Eureka-Server是如何判断一个服务不可用的?

Eureka是通过心跳续约的方式来检查各个服务提供者的健康状态。

实际上,在判断服务不可用这个部分,会分为两块逻辑。

  1. Eureka-Server需要定期检查服务提供者的健康状态。
  2. Eureka-Client在运行过程中需要定期更新注册信息。

Eureka的心跳续约机制如下图所示。

image-20211120163209820
  1. 客户端在启动时, 会开启一个心跳任务,每隔30s向服务单发送一次心跳请求。
  2. 服务端维护了每个实例的最后一次心跳时间,客户端发送心跳包过来后,会更新这个心跳时间。
  3. 服务端在启动时,开启了一个定时任务,该任务每隔60s执行一次,检查每个实例的最后一次心跳时间是否超过90s,如果超过则认为过期,需要剔除。

关于上述流程中涉及到的时间,可以通过以下配置来更改.

#Server 至上一次收到 Client 的心跳之后,等待下一次心跳的超时时间,在这个时间内若没收到下一次心跳,则将移除该 Instance。
eureka.instance.lease-expiration-duration-in-seconds=90
# Server 清理无效节点的时间间隔,默认60000毫秒,即60秒。
eureka.server.eviction-interval-timer-in-ms=60

客户端心跳发起流程

心跳续约是客户端发起的,每隔30s执行一次。

DiscoveryClient.initScheduledTasks

继续回到DiscoveryClient.initScheduledTasks方法中,

private void initScheduledTasks() {
    //省略....
    heartbeatTask = new TimedSupervisorTask(
        "heartbeat",
        scheduler,
        heartbeatExecutor,
        renewalIntervalInSecs,
        TimeUnit.SECONDS,
        expBackOffBound,
        new HeartbeatThread()
    );
    scheduler.schedule(
        heartbeatTask,
        renewalIntervalInSecs, TimeUnit.SECONDS);
    //省略....
}

renewalIntervalInSecs=30s, 默认每隔30s执行一次。

HeartbeatThread

这个线程的实现很简单,调用renew()续约,如果续约成功,则更新最后一次心跳续约时间。

private class HeartbeatThread implements Runnable {

    public void run() {
        if (renew()) {
            lastSuccessfulHeartbeatTimestamp = System.currentTimeMillis();
        }
    }
}

renew()方法中,调用EurekaServer的"apps/" + appName + '/' + id;这个地址,进行心跳续约。

boolean renew() {
    EurekaHttpResponse<InstanceInfo> httpResponse;
    try {
        httpResponse = eurekaTransport.registrationClient.sendHeartBeat(instanceInfo.getAppName(), instanceInfo.getId(), instanceInfo, null);
        logger.debug(PREFIX + "{} - Heartbeat status: {}", appPathIdentifier, httpResponse.getStatusCode());
        if (httpResponse.getStatusCode() == Status.NOT_FOUND.getStatusCode()) {
            REREGISTER_COUNTER.increment();
            logger.info(PREFIX + "{} - Re-registering apps/{}", appPathIdentifier, instanceInfo.getAppName());
            long timestamp = instanceInfo.setIsDirtyWithTime();
            boolean success = register();
            if (success) {
                instanceInfo.unsetIsDirty(timestamp);
            }
            return success;
        }
        return httpResponse.getStatusCode() == Status.OK.getStatusCode();
    } catch (Throwable e) {
        logger.error(PREFIX + "{} - was unable to send heartbeat!", appPathIdentifier, e);
        return false;
    }
}

服务端收到心跳处理

服务端具体为调用[com.netflix.eureka.resources]包下的InstanceResource类的renewLease方法进行续约,代码如下

@PUT
public Response renewLease(
        @HeaderParam(PeerEurekaNode.HEADER_REPLICATION) String isReplication,
        @QueryParam("overriddenstatus") String overriddenStatus,
        @QueryParam("status") String status,
        @QueryParam("lastDirtyTimestamp") String lastDirtyTimestamp) {
    boolean isFromReplicaNode = "true".equals(isReplication);
    //调用renew进行续约
    boolean isSuccess = registry.renew(app.getName(), id, isFromReplicaNode);

    // Not found in the registry, immediately ask for a register
    if (!isSuccess) { //如果续约失败,返回异常
        logger.warn("Not Found (Renew): {} - {}", app.getName(), id);
        return Response.status(Status.NOT_FOUND).build();
    }
    // Check if we need to sync based on dirty time stamp, the client
    // instance might have changed some value
    Response response;
    //校验客户端与服务端的时间差异,如果存在问题则需要重新发起注册
    if (lastDirtyTimestamp != null && serverConfig.shouldSyncWhenTimestampDiffers()) {
        response = this.validateDirtyTimestamp(Long.valueOf(lastDirtyTimestamp), isFromReplicaNode);
        // Store the overridden status since the validation found out the node that replicates wins
        if (response.getStatus() == Response.Status.NOT_FOUND.getStatusCode()
                && (overriddenStatus != null)
                && !(InstanceStatus.UNKNOWN.name().equals(overriddenStatus))
                && isFromReplicaNode) {
            registry.storeOverriddenStatusIfRequired(app.getAppName(), id, InstanceStatus.valueOf(overriddenStatus));
        }
    } else {
        response = Response.ok().build(); // 续约成功,返回200
    }
    logger.debug("Found (Renew): {} - {}; reply status={}", app.getName(), id, response.getStatus());
    return response;
}

InstanceRegistry.renew

renew的实现方法如下,主要有两个流程

  1. 从服务注册列表中找到匹配当前请求的实例
  2. 发布EurekaInstanceRenewedEvent事件
@Override
public boolean renew(final String appName, final String serverId,
                     boolean isReplication) {
    log("renew " + appName + " serverId " + serverId + ", isReplication {}"
        + isReplication);
    //获取所有服务注册信息
    List<Application> applications = getSortedApplications();
    for (Application input : applications) { //逐一遍历
        if (input.getName().equals(appName)) { //如果当前续约的客户端和某个服务注册信息节点相同
            InstanceInfo instance = null;
            for (InstanceInfo info : input.getInstances()) { //遍历这个服务集群下的所有节点,找到某个匹配的实例instance返回。
                if (info.getId().equals(serverId)) {
                    instance = info; //
                    break;
                }
            }
            //发布EurekaInstanceRenewedEvent事件,这个事件在EurekaServer中并没有处理,我们可以监听这个事件来做一些事情,比如做监控。
            publishEvent(new EurekaInstanceRenewedEvent(this, appName, serverId,
                                                        instance, isReplication));
            break;
        }
    }
    return super.renew(appName, serverId, isReplication);
}

super.renew

public boolean renew(final String appName, final String id, final boolean isReplication) {
    if (super.renew(appName, id, isReplication)) { //调用父类的续约方法,如果续约成功
        replicateToPeers(Action.Heartbeat, appName, id, null, null, isReplication); //同步给集群中的所有节点
        return true;
    }
    return false;
}

AbstractInstanceRegistry.renew

在这个方法中,会拿到应用对应的实例列表,然后调用Lease.renew()去进行心跳续约。

public boolean renew(String appName, String id, boolean isReplication) {
    RENEW.increment(isReplication);
    Map<String, Lease<InstanceInfo>> gMap = registry.get(appName); //根据服务名字获取实例信息
    Lease<InstanceInfo> leaseToRenew = null;
    if (gMap != null) { 
        leaseToRenew = gMap.get(id);  //获取需要续约的服务实例,
    }
    if (leaseToRenew == null) { //如果为空,说明这个服务实例不存在,直接返回续约失败
        RENEW_NOT_FOUND.increment(isReplication);
        logger.warn("DS: Registry: lease doesn't exist, registering resource: {} - {}", appName, id);
        return false;
    } else { //表示实例存在
        InstanceInfo instanceInfo = leaseToRenew.getHolder(); //获取实例的基本信息
        if (instanceInfo != null) { //实例基本信息不为空
            // touchASGCache(instanceInfo.getASGName());
            //获取实例的运行状态
            InstanceStatus overriddenInstanceStatus = this.getOverriddenInstanceStatus(
                    instanceInfo, leaseToRenew, isReplication);
            if (overriddenInstanceStatus == InstanceStatus.UNKNOWN) { //如果运行状态未知,也返回续约失败
                logger.info("Instance status UNKNOWN possibly due to deleted override for instance {}"
                        + "; re-register required", instanceInfo.getId());
                RENEW_NOT_FOUND.increment(isReplication);
                return false;
            }
            //如果当前请求的实例信息
            if (!instanceInfo.getStatus().equals(overriddenInstanceStatus)) {
                logger.info(
                        "The instance status {} is different from overridden instance status {} for instance {}. "
                                + "Hence setting the status to overridden status", instanceInfo.getStatus().name(),
                                overriddenInstanceStatus.name(),
                                instanceInfo.getId());
                instanceInfo.setStatusWithoutDirty(overriddenInstanceStatus);

            }
        }
        //更新上一分钟的续约数量
        renewsLastMin.increment();
        leaseToRenew.renew(); //续约
        return true;
    }
}

续约的实现,就是更新服务端最后一次收到心跳请求的时间。

public void renew() {
    lastUpdateTimestamp = System.currentTimeMillis() + duration;

}

Eureka的自我保护机制

实际,心跳检测机制有一定的不确定行,比如服务提供者可能是正常的,但是由于网络通信的问题,导致在90s内没有收到心跳请求,那将会导致健康的服务被误杀。

为了避免这种问题,Eureka提供了一种叫自我保护机制的东西。简单来说,就是开启自我保护机制后,Eureka Server会包这些服务实例保护起来,避免过期导致实例被剔除的问题,从而保证Eurreka集群更加健壮和稳定。

进入自我保护状态后,会出现以下几种情况

Eureka自我保护机制,通过配置 eureka.server.enable-self-preservation 来【true】打开/【false禁用】自我保护机制,默认打开状态,建议生产环境打开此配置。

自我保护机制应该如何设计,才能更加精准的控制到“是网络异常”导致的通信延迟,而不是服务宕机呢?

Eureka是这么做的: 如果低于85%的客户端节点都没有正常的心跳,那么Eureka Server就认为客户端与注册中心出现了网络故障,Eureka Server自动进入自我保护状态.

其中,85%这个阈值,可以通过下面这个配置来设置

# 自我保护续约百分比,默认是0.85
eureka.server.renewal-percent-threshold=0.85

但是还有个问题,超过谁的85%呢?这里有一个预期的续约数量,这个数量计算公式如下:

//自我保护阀值 = 服务总数 * 每分钟续约数(60S/客户端续约间隔) * 自我保护续约百分比阀值因子

假设如果有100个服务,续约间隔是30S,自我保护阈值0.85,那么它的预期续约数量为:

自我保护阈值 =100 * 60 / 30 * 0.85 = 170。

自动续约的阈值设置

在EurekaServerBootstrap这个类的contextInitialized方法中,会调用initEurekaServerContext进行初始化

public void contextInitialized(ServletContext context) {
    try {
        initEurekaEnvironment();
        initEurekaServerContext();

        context.setAttribute(EurekaServerContext.class.getName(), this.serverContext);
    }
    catch (Throwable e) {
        log.error("Cannot bootstrap eureka server :", e);
        throw new RuntimeException("Cannot bootstrap eureka server :", e);
    }
}

继续往下看。

protected void initEurekaServerContext() throws Exception {
        EurekaServerConfig eurekaServerConfig = new DefaultEurekaServerConfig();
    //...
    registry.openForTraffic(applicationInfoManager, registryCount);
}

在openForTraffic方法中,会初始化expectedNumberOfClientsSendingRenews这个值,这个值的含义是: 预期每分钟收到续约的客户端数量,取决于注册到eureka server上的服务数量

@Override
public void openForTraffic(ApplicationInfoManager applicationInfoManager, int count) {
    // Renewals happen every 30 seconds and for a minute it should be a factor of 2.
    this.expectedNumberOfClientsSendingRenews = count; //初始值是1.
    updateRenewsPerMinThreshold();
    logger.info("Got {} instances from neighboring DS node", count);
    logger.info("Renew threshold is: {}", numberOfRenewsPerMinThreshold);
    this.startupTime = System.currentTimeMillis();
    if (count > 0) {
        this.peerInstancesTransferEmptyOnStartup = false;
    }
    DataCenterInfo.Name selfName = applicationInfoManager.getInfo().getDataCenterInfo().getName();
    boolean isAws = Name.Amazon == selfName;
    if (isAws && serverConfig.shouldPrimeAwsReplicaConnections()) {
        logger.info("Priming AWS connections for all replicas..");
        primeAwsReplicas(applicationInfoManager);
    }
    logger.info("Changing status to UP");
    applicationInfoManager.setInstanceStatus(InstanceStatus.UP);
    super.postInit();
}

updateRenewsPerMinThreshold

接着调用updateRenewsPerMinThreshold方法,会更新一个每分钟最小的续约数量,也就是Eureka Server期望每分钟收到客户端实例续约的总数的阈值。如果小于这个阈值,就会触发自我保护机制。

protected void updateRenewsPerMinThreshold() {
    this.numberOfRenewsPerMinThreshold = (int) (this.expectedNumberOfClientsSendingRenews
            * (60.0 / serverConfig.getExpectedClientRenewalIntervalSeconds())
            * serverConfig.getRenewalPercentThreshold());
}
//自我保护阀值 = 服务总数 * 每分钟续约数(60S/客户端续约间隔) * 自我保护续约百分比阀值因子

预期值的变化触发机制

expectedNumberOfClientsSendingRenewsnumberOfRenewsPerMinThreshold 这两个值,会随着新增服务注册以及服务下线的触发而发生变化。

PeerAwareInstanceRegistryImpl.cancel

当服务提供者主动下线时,表示这个时候Eureka-Server要剔除这个服务提供者的地址,同时也代表这这个心跳续约的阈值要发生变化。所以在PeerAwareInstanceRegistryImpl.cancel中可以看到数据的更新

调用路径 PeerAwareInstanceRegistryImpl.cancel -> AbstractInstanceRegistry.cancel->internalCancel

服务下线之后,意味着需要发送续约的客户端数量递减了,所以在这里进行修改

protected boolean internalCancel(String appName, String id, boolean isReplication) {
  //....
    synchronized (lock) {
        if (this.expectedNumberOfClientsSendingRenews > 0) {
            // Since the client wants to cancel it, reduce the number of clients to send renews.
            this.expectedNumberOfClientsSendingRenews = this.expectedNumberOfClientsSendingRenews - 1;
            updateRenewsPerMinThreshold();
        }
    }
}

PeerAwareInstanceRegistryImpl.register

当有新的服务提供者注册到eureka-server上时,需要增加续约的客户端数量,所以在register方法中会进行处理

register ->super.register(AbstractInstanceRegistry)

public void register(InstanceInfo registrant, int leaseDuration, boolean isReplication) {
    //....    
    // The lease does not exist and hence it is a new registration
    synchronized (lock) {
        if (this.expectedNumberOfClientsSendingRenews > 0) {
            // Since the client wants to register it, increase the number of clients sending renews
            this.expectedNumberOfClientsSendingRenews = this.expectedNumberOfClientsSendingRenews + 1;
            updateRenewsPerMinThreshold();
        }
    }
}

每隔15分钟刷新自我保护阈值

PeerAwareInstanceRegistryImpl.scheduleRenewalThresholdUpdateTask

每隔15分钟,更新一次自我保护阈值!

private void updateRenewalThreshold() {
    try {
        // 1. 计算应用实例数
        Applications apps = eurekaClient.getApplications();
        int count = 0;
        for (Application app : apps.getRegisteredApplications()) {
            for (InstanceInfo instance : app.getInstances()) {
                if (this.isRegisterable(instance)) {
                    ++count;
                }
            }
        }
        
        synchronized (lock) {
            // Update threshold only if the threshold is greater than the
            // current expected threshold or if self preservation is disabled.
            //当节点数量count大于最小续约数量时,或者没有开启自我保护机制的情况下,重新计算expectedNumberOfClientsSendingRenews和numberOfRenewsPerMinThreshold
            if ((count) > (serverConfig.getRenewalPercentThreshold() * expectedNumberOfClientsSendingRenews)
                || (!this.isSelfPreservationModeEnabled())) {
                this.expectedNumberOfClientsSendingRenews = count;
                updateRenewsPerMinThreshold();
            }
        }
        logger.info("Current renewal threshold is : {}", numberOfRenewsPerMinThreshold);
    } catch (Throwable e) {
        logger.error("Cannot update renewal threshold", e);
    }
}

自我保护机制的触发

AbstractInstanceRegistrypostInit方法中,会开启一个EvictionTask的任务,这个任务用来检测是否需要开启自我保护机制。

这个方法也是在EurekaServerBootstrap方法启动时触发。

protected void postInit() {
    renewsLastMin.start(); //开启一个定时任务,用来实现每分钟的续约数量,每隔60s归0重新计算
    if (evictionTaskRef.get() != null) {
        evictionTaskRef.get().cancel();
    }
    evictionTaskRef.set(new EvictionTask()); //启动一个定时任务EvictionTask,每隔60s执行一次
    evictionTimer.schedule(evictionTaskRef.get(),
                           serverConfig.getEvictionIntervalTimerInMs(),
                           serverConfig.getEvictionIntervalTimerInMs());
}

其中,EvictionTask的代码如下。

private final AtomicLong lastExecutionNanosRef = new AtomicLong(0l);

@Override
public void run() {
    try {
        //获取补偿时间毫秒数
        long compensationTimeMs = getCompensationTimeMs();
        logger.info("Running the evict task with compensationTime {}ms", compensationTimeMs);
        evict(compensationTimeMs);
    } catch (Throwable e) {
        logger.error("Could not run the evict task", e);
    }
}

evict方法

public void evict(long additionalLeaseMs) {
    logger.debug("Running the evict task");
     // 是否需要开启自我保护机制,如果需要,那么直接RETURE, 不需要继续往下执行了
    if (!isLeaseExpirationEnabled()) {
        logger.debug("DS: lease expiration is currently disabled.");
        return;
    }

    //这下面主要是做服务自动下线的操作的。
}

isLeaseExpirationEnabled

public boolean isLeaseExpirationEnabled() {
    if (!isSelfPreservationModeEnabled()) {
        // The self preservation mode is disabled, hence allowing the instances to expire.
        return true;
    }
    return numberOfRenewsPerMinThreshold > 0 && getNumOfRenewsInLastMin() > numberOfRenewsPerMinThreshold;
}
上一篇下一篇

猜你喜欢

热点阅读