摘要: 原创出处 http://www.iocoder.cn/Eureka/instance-registry-self-preservation/ 「芋道源码」欢迎转载,保留摘要,谢谢!
本文主要基于 Eureka 1.8.X 版本
- 1. 概述
- 2. 定义
- 3. 实现
1. 概述
本文主要分享 自我保护机制,为应用实例过期下线做铺垫。
推荐 Spring Cloud 书籍:
- 请支持正版。下载盗版,等于主动编写低级 BUG 。
- 程序猿DD —— 《Spring Cloud微服务实战》
- 周立 —— 《Spring Cloud与Docker微服务架构实战》
2. 定义
自我保护机制定义如下:
FROM 周立 —— 《理解Eureka的自我保护模式》
当Eureka Server节点在短时间内丢失过多客户端时(可能发生了网络分区故障),那么这个节点就会进入自我保护模式。一旦进入该模式,Eureka Server就会保护服务注册表中的信息,不再删除服务注册表中的数据(也就是不会注销任何微服务)。当网络故障恢复后,该Eureka Server节点会自动退出自我保护模式。为什么使用自动保护机制 ?你也可以从周立兄的这篇文章得到答案,这里笔者就不一本正经的胡说八道了。
3. 实现
首先,我们来看下在自动保护机制里扮演重要角色的两个变量:
// AbstractInstanceRegistry.java/* 期望最小每分钟续租次数/protected volatile int numberOfRenewsPerMinThreshold;/期望最大每分钟续租次数*/protected volatile int expectedNumberOfRenewsPerMin;expectedNumberOfRenewsPerMin
,期望最大每分钟续租次数。numberOfRenewsPerMinThreshold
,期望最小每分钟续租次数。
3.1 触发条件
当每分钟心跳次数(
renewsLastMin
) 小于numberOfRenewsPerMinThreshold
时,并且开启自动保护模式开关(eureka.enableSelfPreservation = true
) 时,触发自动保护机制,不再自动过期租约,实现代码如下:
// AbstractInstanceRegistry.javapublic void evict(long additionalLeaseMs) {if (!isLeaseExpirationEnabled()) {logger.debug( "DS: lease expiration is currently disabled.");return;}// ... 省略过期租约逻辑}// PeerAwareInstanceRegistryImpl.javapublic boolean isLeaseExpirationEnabled() {if (!isSelfPreservationModeEnabled()) {// The self preservation mode is disabled, hence allowing the instances to expire.return true;}return numberOfRenewsPerMinThreshold > 0 && getNumOfRenewsInLastMin() > numberOfRenewsPerMinThreshold;}
3.2 计算公式
计算公式如下:
expectedNumberOfRenewsPerMin
= 当前注册的应用实例数x
2numberOfRenewsPerMinThreshold
=expectedNumberOfRenewsPerMin
*
续租百分比(eureka.renewalPercentThreshold
)
为什么乘以 2
默认情况下,注册的应用实例每半分钟续租一次,那么一分钟心跳两次,因此 x 2 。
这块会有一些硬编码的情况,因此不太建议修改应用实例的续租频率。
为什么乘以续租百分比
低于这个百分比,意味着开启自我保护机制。
默认情况下,
eureka.renewalPercentThreshold = 0.85
。如果你真的调整了续租频率,可以等比去续租百分比,以保证合适的触发自我保护机制的阀值。另外,你需要注意,续租频率是 Client 级别,续租百分比是 Server 级别。
3.3 计算时机
目前有四个地方会计算
numberOfRenewsPerMinThreshold
、expectedNumberOfRenewsPerMin
,我们逐小节来看。
3.3.1 Eureka-Server 初始化
Eureka-Server 在启动时,从 Eureka-Server 集群获取注册信息,并首次初始化
numberOfRenewsPerMinThreshold
、expectedNumberOfRenewsPerMin
。实现代码如下:
// EurekaBootStrap.javaprotected void initEurekaServerContext() throws Exception {// ... 省略其它代码// 【2.2.10】从其他 Eureka-Server 拉取注册信息// Copy registry from neighboring eureka nodeint registryCount = registry.syncUp();registry.openForTraffic(applicationInfoManager, registryCount);// ... 省略其它代码}// PeerAwareInstanceRegistryImpl.javapublic void openForTraffic(ApplicationInfoManager applicationInfoManager, int count) {// Renewals happen every 30 seconds and for a minute it should be a factor of 2.this.expectedNumberOfRenewsPerMin = count * 2;this.numberOfRenewsPerMinThreshold =( int) ( this.expectedNumberOfRenewsPerMin * serverConfig.getRenewalPercentThreshold());// ... 省略其它代码}
3.3.2 定时重置
Eureka-Server 定时重新计算
numberOfRenewsPerMinThreshold
、expectedNumberOfRenewsPerMin
。实现代码如下:
// PeerAwareInstanceRegistryImpl.javaprivate void scheduleRenewalThresholdUpdateTask() {timer.schedule( new TimerTask() {public void run() {updateRenewalThreshold();}}, serverConfig.getRenewalThresholdUpdateIntervalMs(),serverConfig.getRenewalThresholdUpdateIntervalMs());}// AbstractInstanceRegistry.java/*自我保护机锁当计算如下参数时使用:* 1. {@link #numberOfRenewsPerMinThreshold}* 2. {@link #expectedNumberOfRenewsPerMin}/protected final Object lock = new Object();private void updateRenewalThreshold() {try {// 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;}}}// 计算 expectedNumberOfRenewsPerMin 、 numberOfRenewsPerMinThreshold 参数synchronized (lock) {// Update threshold only if the threshold is greater than the// current expected threshold of if the self preservation is disabled.if ((count 2) > (serverConfig.getRenewalPercentThreshold() * numberOfRenewsPerMinThreshold)|| (! this.isSelfPreservationModeEnabled())) {this.expectedNumberOfRenewsPerMin = count * 2;this.numberOfRenewsPerMinThreshold = ( int) ((count * 2) * serverConfig.getRenewalPercentThreshold());}}logger.info( "Current renewal threshold is : {}", numberOfRenewsPerMinThreshold);} catch (Throwable e) {logger.error( "Cannot update renewal threshold", e);}}- 配置
eureka.renewalThresholdUpdateIntervalMs
参数,定时重新计算。默认,15 分钟。 - 代码块
!this.isSelfPreservationModeEnabled()
:当未开启自我保护机制时,每次都进行重新计算。事实上,这两个参数不仅仅自我保护机制会使用到,配合 Netflix Servo 实现监控信息采集numberOfRenewsPerMinThreshold
、expectedNumberOfRenewsPerMin
。 - 代码块
(count * 2) > (serverConfig.getRenewalPercentThreshold() * numberOfRenewsPerMinThreshold)
:当开启自我保护机制时,应用实例每分钟最大心跳数(count * 2
) 小于期望最小每分钟续租次数(serverConfig.getRenewalPercentThreshold() * numberOfRenewsPerMinThreshold
),不重新计算。如果重新计算,自动保护机制会每次定时执行后失效。
3.3.3 应用实例注册
应用实例注册时,增加
numberOfRenewsPerMinThreshold
、expectedNumberOfRenewsPerMin
。实现代码如下:
//public void register(InstanceInfo registrant, int leaseDuration, boolean isReplication) {// ... 省略无关代码// The lease does not exist and hence it is a new registration// 【自我保护机制】增加numberOfRenewsPerMinThreshold
、expectedNumberOfRenewsPerMin
synchronized (lock) {if ( this.expectedNumberOfRenewsPerMin > 0) {// Since the client wants to cancel it, reduce the threshold// (1// for 30 seconds, 2 for a minute)this.expectedNumberOfRenewsPerMin = this.expectedNumberOfRenewsPerMin + 2;this.numberOfRenewsPerMinThreshold =( int) ( this.expectedNumberOfRenewsPerMin * serverConfig.getRenewalPercentThreshold());}}// ... 省略无关代码}
3.3.4 应用实例下线
应用实例下线时,减少
numberOfRenewsPerMinThreshold
、expectedNumberOfRenewsPerMin
。实现代码如下:
// PeerAwareInstanceRegistryImpl.javapublic boolean cancel(final String appName, final String id,final boolean isReplication) {// ... 省略无关代码synchronized (lock) {if ( this.expectedNumberOfRenewsPerMin > 0) {// Since the client wants to cancel it, reduce the threshold (1 for 30 seconds, 2 for a minute)this.expectedNumberOfRenewsPerMin = this.expectedNumberOfRenewsPerMin - 2;this.numberOfRenewsPerMinThreshold = ( int) ( this.expectedNumberOfRenewsPerMin * serverConfig.getRenewalPercentThreshold());}}// ... 省略无关代码}