理解eureka的自我保护机制

计算公式

Number of heartbeats expected from one client instance / min

factor &＃61; 60/leaseRenewalIntervalInSeconds

Number of heartbeats expected from N instances / min

factor * N

Minimum expected heartbeat threshold / min

factor N renewalPercentThreshold

源代码

this.expectedNumberOfRenewsPerMin &＃61; N * 2; this.numberOfRenewsPerMinThreshold &＃61; (int) (this.expectedNumberOfRenewsPerMin * serverConfig.getRenewalPercentThreshold());

eureka启动时

eureka-core-1.4.12-sources.jar!/com/netflix/eureka/registry/PeerAwareInstanceRegistryImpl.java

&＃64;Overridepublic void openForTraffic(ApplicationInfoManager applicationInfoManager, int count) {// Renewals happen every 30 seconds and for a minute it should be a factor of 2.this.expectedNumberOfRenewsPerMin &＃61; count * 2;this.numberOfRenewsPerMinThreshold &＃61;(int) (this.expectedNumberOfRenewsPerMin * serverConfig.getRenewalPercentThreshold());logger.info("Got " &＃43; count &＃43; " instances from neighboring DS node");logger.info("Renew threshold is: " &＃43; numberOfRenewsPerMinThreshold);this.startupTime &＃61; System.currentTimeMillis();if (count > 0) {this.peerInstancesTransferEmptyOnStartup &＃61; false;}DataCenterInfo.Name selfName &＃61; applicationInfoManager.getInfo().getDataCenterInfo().getName();boolean isAws &＃61; Name.Amazon &＃61;&＃61; 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();}

count为eureka个数&＃xff0c;此时expectedNumberOfRenewsPerMin为2(1个注册中心&＃xff0c;每30秒发送一次心跳&＃xff0c;则每分钟对于每个client&＃xff0c;应该发送2次心跳)&＃xff0c;numberOfRenewsPerMinThreshold&＃61;&＃xff08;int&＃xff09;2*0.85 &＃61;1

client注册时

eureka-core-1.4.12-sources.jar!/com/netflix/eureka/registry/AbstractInstanceRegistry.java

public void register(InstanceInfo registrant, int leaseDuration, boolean isReplication) {try {read.lock();Map> gMap &＃61; registry.get(registrant.getAppName());REGISTER.increment(isReplication);if (gMap &＃61;&＃61; null) {final ConcurrentHashMap> gNewMap &＃61; new ConcurrentHashMap>();gMap &＃61; registry.putIfAbsent(registrant.getAppName(), gNewMap);if (gMap &＃61;&＃61; null) {gMap &＃61; gNewMap;}}Lease existingLease &＃61; gMap.get(registrant.getId());// Retain the last dirty timestamp without overwriting it, if there is already a leaseif (existingLease !&＃61; null && (existingLease.getHolder() !&＃61; null)) {Long existingLastDirtyTimestamp &＃61; existingLease.getHolder().getLastDirtyTimestamp();Long registrationLastDirtyTimestamp &＃61; registrant.getLastDirtyTimestamp();logger.debug("Existing lease found (existing&＃61;{}, provided&＃61;{}", existingLastDirtyTimestamp, registrationLastDirtyTimestamp);if (existingLastDirtyTimestamp > registrationLastDirtyTimestamp) {logger.warn("There is an existing lease and the existing lease&＃39;s dirty timestamp {} is greater" &＃43;" than the one that is being registered {}", existingLastDirtyTimestamp, registrationLastDirtyTimestamp);logger.warn("Using the existing instanceInfo instead of the new instanceInfo as the registrant");registrant &＃61; existingLease.getHolder();}} else {// The lease does not exist and hence it is a new registrationsynchronized (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 &＃61; this.expectedNumberOfRenewsPerMin &＃43; 2;this.numberOfRenewsPerMinThreshold &＃61;(int) (this.expectedNumberOfRenewsPerMin * serverConfig.getRenewalPercentThreshold());}}logger.debug("No previous lease information found; it is new registration");}Lease lease &＃61; new Lease(registrant, leaseDuration);if (existingLease !&＃61; null) {lease.setServiceUpTimestamp(existingLease.getServiceUpTimestamp());}gMap.put(registrant.getId(), lease);synchronized (recentRegisteredQueue) {recentRegisteredQueue.add(new Pair(System.currentTimeMillis(),registrant.getAppName() &＃43; "(" &＃43; registrant.getId() &＃43; ")"));}// This is where the initial state transfer of overridden status happensif (!InstanceStatus.UNKNOWN.equals(registrant.getOverriddenStatus())) {logger.debug("Found overridden status {} for instance {}. Checking to see if needs to be add to the "&＃43; "overrides", registrant.getOverriddenStatus(), registrant.getId());if (!overriddenInstanceStatusMap.containsKey(registrant.getId())) {logger.info("Not found overridden id {} and hence adding it", registrant.getId());overriddenInstanceStatusMap.put(registrant.getId(), registrant.getOverriddenStatus());}}InstanceStatus overriddenStatusFromMap &＃61; overriddenInstanceStatusMap.get(registrant.getId());if (overriddenStatusFromMap !&＃61; null) {logger.info("Storing overridden status {} from map", overriddenStatusFromMap);registrant.setOverriddenStatus(overriddenStatusFromMap);}// Set the status based on the overridden status rulesInstanceStatus overriddenInstanceStatus &＃61; getOverriddenInstanceStatus(registrant, existingLease, isReplication);registrant.setStatusWithoutDirty(overriddenInstanceStatus);// If the lease is registered with UP status, set lease service up timestampif (InstanceStatus.UP.equals(registrant.getStatus())) {lease.serviceUp();}registrant.setActionType(ActionType.ADDED);recentlyChangedQueue.add(new RecentlyChangedItem(lease));registrant.setLastUpdatedTimestamp();invalidateCache(registrant.getAppName(), registrant.getVIPAddress(), registrant.getSecureVipAddress());logger.info("Registered instance {}/{} with status {} (replication&＃61;{})",registrant.getAppName(), registrant.getId(), registrant.getStatus(), isReplication);} finally {read.unlock();}}

其中&＃xff0c;重点看这段

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 &＃61; this.expectedNumberOfRenewsPerMin &＃43; 2;this.numberOfRenewsPerMinThreshold &＃61;(int) (this.expectedNumberOfRenewsPerMin * serverConfig.getRenewalPercentThreshold());}}

每注册上一个实例&＃xff0c;重新算一下&＃xff0c;即

this.expectedNumberOfRenewsPerMin &＃61; this.expectedNumberOfRenewsPerMin &＃43; 2; this.numberOfRenewsPerMinThreshold &＃61; (int) (this.expectedNumberOfRenewsPerMin * serverConfig.getRenewalPercentThreshold());

总的来说&＃xff0c;所有实例都注册上了&＃xff0c;就是

this.expectedNumberOfRenewsPerMin &＃61; N * 2; this.numberOfRenewsPerMinThreshold &＃61; (int) (this.expectedNumberOfRenewsPerMin * serverConfig.getRenewalPercentThreshold());

计算例子

2个eureka,算在内,factor&＃61;6,N:16

leaseRenewalIntervalInSeconds: 10 --> 每分钟6次 this.expectedNumberOfRenewsPerMin &＃61; N * 2 &＃61; 32 ; this.numberOfRenewsPerMinThreshold &＃61; (int) (32 * 0.85) &＃61; (int)27.2 &＃61; 27;

2个eureka,算在内,factor&＃61;2,N&＃61;3

leaseRenewalIntervalInSeconds: 30 --> 每分钟2次 this.expectedNumberOfRenewsPerMin &＃61; N * 2 &＃61; 6 ; this.numberOfRenewsPerMinThreshold &＃61; (int) (6* 0.85) &＃61; (int)5.1 &＃61; 5;

对于单机版eureka的&＃xff0c;比较特殊&＃xff0c;由于openForTraffic这个方法里头

this.expectedNumberOfRenewsPerMin &＃61; count * 2;

这里count&＃61;1多算了一个实例的心跳次数&＃xff0c;所以如下&＃xff1a;

1个eureka,不注册,factor&＃61;2,N&＃61;1

leaseRenewalIntervalInSeconds: 30 --> 每分钟2次 this.expectedNumberOfRenewsPerMin &＃61; N * 2 &＃61; 2 ; this.numberOfRenewsPerMinThreshold &＃61; (int) (2 * 0.85) &＃61; (int)1.7 &＃61; 1;实际显示是3&＃xff0c;应该是按N&＃61;2算 this.expectedNumberOfRenewsPerMin &＃61; N * 2 &＃61; 4 ; this.numberOfRenewsPerMinThreshold &＃61; (int) (4 * 0.85) &＃61; (int)3.4 &＃61; 3;

1个eureka,不注册,factor&＃61;2,N&＃61;2

leaseRenewalIntervalInSeconds: 30 --> 每分钟2次 this.expectedNumberOfRenewsPerMin &＃61; N * 2 &＃61; 4 ; this.numberOfRenewsPerMinThreshold &＃61; (int) (4 * 0.85) &＃61; (int)3.4 &＃61; 3; 实际显示是5&＃xff0c;应该是按N&＃61;3算 this.expectedNumberOfRenewsPerMin &＃61; N * 2 &＃61; 6 ; this.numberOfRenewsPerMinThreshold &＃61; (int) (6 * 0.85) &＃61; (int)5.1 &＃61; 5;

反思源码

由于eureka-core-1.4.12版本里头&＃xff0c;你去调整eureka.instance.leaseRenewalIntervalInSeconds的话&＃xff0c;代码里头没有相应调整factor&＃xff0c;也就是代码还是60/30&＃61;2&＃xff0c;所以会破坏eureka内置的设计思路。不过对于小型项目来说&＃xff0c;没有跨机房&＃xff0c;网络没有那么恶劣的话&＃xff0c;想避免自我保护导致的服务注册列表不能修改的问题&＃xff0c;可以选择以下任一方式尝试下&＃xff1a;

关闭自我保护eureka.server.enableSelfPreservation&＃61;false
调小eureka.instance.leaseRenewalIntervalInSeconds&＃xff0c;比如设置为10秒
调小renewalPercentThreshold&＃xff0c;比如改为0.49

另外&＃xff0c;还有一个参数可以调整&＃xff0c;就是心跳阈值重新计算的周期&＃xff1a;
eureka.server.renewalThresholdUpdateIntervalMs &＃61; 15 60 1000 # 15mins
默认是15分钟可以改小一点&＃xff0c;比如5分钟&＃61;5601000