sprincloud源码之eureka服务注册
前言eureka是netflix公司基于jersey框架写的一个注册中心,提供了服务注册,服务下架,服务续约,集群同步等功能。 jersey是一个类似于springmvc的框架,只不过mvc是基于servlet的,jersey是基于filter的,二者在使用上也很类似,mvc发请求被servlet拦截到反射调用controller,而jersey是被filter拦截到调用resource, 二者的原理基本一致。
sprincloud整合eureka@EnableEurekaServer @EnableEurekaServer----->>
@Import(EurekaServerMarkerConfiguration.class)
public @interface EnableEurekaServer {
}
----->>>
@Configuration(proxyBeanMethods = false)
public class EurekaServerMarkerConfiguration {
@Bean
public Marker eurekaServerMarkerBean() {
return new Marker();
}
class Marker {
}
}
上面代码的意思就是加了@EnableEurekaServer就会往spring容器注入一个Marker对象 EurekaServerAutoConfiguration spring.factories里的内容如下 org.springframework.boot.autoconfigure.EnableAutoConfiguration= org.springframework.cloud.netflix.eureka.server.EurekaServerAutoConfiguration
这是springboot自动配置的原理,springboot在初始化的时候会把spring.factories定义的类也初始化,可认为是spi @ConditionalOnBean(EurekaServerMarkerConfiguration.Marker.class)
public class EurekaServerAutoConfiguration
这里我把其他注解省略了,可以看到EurekaServerAutoConfiguration初始化的前提是spring容器得有Marker类,
也就是说加上@EnableEurekaServer就是eureka服务注册中心
jersey过滤器的初始化 springmvc自动配置的时候是在DispatcherServletAutoConfiguration初始化DispatcherServlet, 现在也是如此,在EurekaServerAutoConfiguration初始化一个filter,代码如下 @Bean
public FilterRegistrationBean<?> jerseyFilterRegistration(
javax.ws.rs.core.Application eurekaJerseyApp) {
FilterRegistrationBean<Filter> bean = new FilterRegistrationBean<Filter>();
bean.setFilter(new ServletContainer(eurekaJerseyApp));
bean.setOrder(Ordered.LOWEST_PRECEDENCE);
bean.setUrlPatterns(
Collections.singletonList(EurekaConstants.DEFAULT_PREFIX + "/*"));
return bean;
}
filter初始化后就可以拦截eureka-client的请求转发给eureka-server处理了
eureka服务注册源码在ApplicationResource的addInstance方法
ApplicationResource.addInstance()---->>
this.registry.register()---->>
InstanceRegistry.register(){
//springcloud发布一个事件EurekaInstanceRegisteredEvent
handleRegistration(info, resolveInstanceLeaseDuration(info), isReplication);
//调用父类的register
super.register(info, isReplication);
}
我们先暂停一下如上代码讲解,先看看InstanceRegistry的继承关系图:
InstanceRegistry是springcloud对于eureka的一个扩展,此类的唯一作用就是发布事件 PeerAwareInstanceRegistrImpl是专门做集群同步的 AbstractInstanceRegistry才是做服务注册的 这里采取了一种责任链模式,先发布事件,在服务注册最后集群同步,每一个类只做一件事 下面我们接着代码说: InstanceRegistry.register 发布事件 InstanceRegistry.register(){
//springcloud发布一个事件EurekaInstanceRegisteredEvent
handleRegistration(info, resolveInstanceLeaseDuration(info), isReplication);
//调用父类的register
super.register(info, isReplication);
}
PeerAwareInstanceRegistrImpl.register public void register(InstanceInfo info, boolean isReplication) {
//leaseExpirationDurationInSeconds: 30 #Eureka服务器在接收到实例的最后一次发出的心跳后,需要等待多久才可以将此实例删除,默认为90秒
int leaseDuration = 90;
if (info.getLeaseInfo() != null && info.getLeaseInfo().getDurationInSecs() > 0) {
leaseDuration = info.getLeaseInfo().getDurationInSecs();
}
//服务注册
super.register(info, leaseDuration, isReplication);
//集群同步
this.replicateToPeers(PeerAwareInstanceRegistryImpl.Action.Register, info.getAppName(), info.getId(), info, (InstanceStatus)null, isReplication);
}
AbstractInstanceRegistry.register 服务注册 在看服务注册的代码前我们先看this.registry和租债器Lease this.registry ConcurrentHashMap<String, Map<String, Lease<InstanceInfo>>> registry = new ConcurrentHashMap();
registry的数据结构如上等价于<applicationName, Map<instanceId, Lease>> 如图所示有三个order微服务,applicationName配置的都是order,说明三个微服务是属于同一个微服务组,而三个微服务的id各不相同,代表这是同一个order组的三个微服务,这样user微服务来调用order模块就知道如何负载均衡了。 租债器Lease 注释的代码是我修正netflix的代码
public class Lease<T> {
//Eureka服务器在接收到实例的最后一次发出的心跳后,需要等待多久才可以将此实例删除,默认为90秒
public static final int DEFAULT_DURATION_IN_SECS = 90;
private T holder;//微服务对象
private long evictionTimestamp;//服务剔除时间戳
private long registrationTimestamp;//服务注册时间戳
private long serviceUpTimestamp;//
private volatile long lastUpdateTimestamp;//最后一次更新的时间,注册下架续约都是更新操作
private long duration;//duration ms后没有心跳剔除服务
//private long expireTimestamp;//过期时间
/**
* @param r 微服务实例
* @param durationInSecs Eureka服务器在接收到实例的最后一次发出的心跳后,需要等待多久才可以将此实例删除
* eureka.instance.leaseExpirationDurationInSeconds=30s,后一次发出的心跳后,30s后还没有新的心跳剔除服务
* eureka.instance.leaseRenewalIntervalInSeconds=10s心跳间隔时间
*/
public Lease(T r, int durationInSecs) {
this.holder = r;
this.registrationTimestamp = System.currentTimeMillis();//注册时间
this.lastUpdateTimestamp = this.registrationTimestamp;//更新时间=注册时间
this.duration = (long)(durationInSecs * 1000);//duration ms后没有心跳剔除服务
}
//续期操作
public void renew() {
this.lastUpdateTimestamp = System.currentTimeMillis() + this.duration;
// this.lastUpdateTimestamp = System.currentTimeMillis();
// this.expireTimestamp = System.currentTimeMillis() + this.duration;
}
//服务下架/剔除操作
public void cancel() {
if (this.evictionTimestamp <= 0L) {
this.evictionTimestamp = System.currentTimeMillis();
}
}
public void serviceUp() {
if (this.serviceUpTimestamp == 0L) {
this.serviceUpTimestamp = System.currentTimeMillis();
}
}
public boolean isExpired() {
return this.isExpired(0L);
}
public boolean isExpired(long additionalLeaseMs) {
return this.evictionTimestamp > 0L || System.currentTimeMillis() > this.lastUpdateTimestamp + this.duration + additionalLeaseMs;
}
// public boolean isExpired(long additionalLeaseMs) {
// return this.evictionTimestamp > 0L || System.currentTimeMillis() > this.expireTimestamp + additionalLeaseMs;
// }
public static void main(String[] args) throws InterruptedException {
Lease lease = new Lease(new User(1,"lry"),5);
while(true){
System.out.println(lease.isExpired());
Thread.sleep(1000);
}
}
static class User{
private int id;
private String name;
public User(int id,String name){
this.id = id;
this.name = name;
}
}
}
上面的代码netflix公司表示有bug,具体是如下两个方法 //续期
//把最后更新时间改为当前时间+duration
public void renew() {
lastUpdateTimestamp = System.currentTimeMillis() + duration;
}
//对象是否过期
//System.currentTimeMillis() > lastUpdateTimestamp + duration
//判断是否过期的时候又加了一个duration 相当于是2*duration 后才算是过期
public boolean isExpired(long additionalLeaseMs) {
return (evictionTimestamp > 0 || System.currentTimeMillis() > (lastUpdateTimestamp + duration + additionalLeaseMs));
}
bug描述 Note that due to renew() doing the 'wrong" thing and setting lastUpdateTimestamp to +duration
more than what it should be, the expiry will actually be 2 * duration. This is a minor bug and
should only affect instances that ungracefully shutdown. Due to possible wide ranging impact
to existing usage, this will not be fixed.
大致意思就是renew多加了一个duration导致判断isExpire实际是2*duration
服务注册代码 public void register(InstanceInfo registrant, int leaseDuration, boolean isReplication) {
//Lease可以理解成一个微服务实例,它内部封装一个微服务和各种时间来实现对对象的过期与否控制
//可以理解lease==InstanceInfo
Map<String, Lease<InstanceInfo>> gMap = (Map)this.registry.get(registrant.getAppName());
//registrant.getAppName()微服务组内没有微服务
if (gMap == null) {
ConcurrentHashMap<String, Lease<InstanceInfo>> gNewMap = new ConcurrentHashMap();
gMap = (Map)this.registry.putIfAbsent(registrant.getAppName(), gNewMap);
if (gMap == null) {
gMap = gNewMap;
}
}
//尝试通过id拿到一个微服务实例,一般情况下都拿不到,除非以下两种情况
//情况一:有两台application name 和instance id都一样微服务
//情况二:断点调试,这种情况冲突是因为客户端注册有超时重试机制
Lease<InstanceInfo> existingLease = gMap.get(registrant.getId());
//如果拿到了
if (existingLease != null && (existingLease.getHolder() != null)) {
//已经存在的微服务实例最后修改时间
Long existingLastDirtyTimestamp = existingLease.getHolder().getLastDirtyTimestamp();
//要注册的微服务实例最后修改时间
Long registrationLastDirtyTimestamp = registrant.getLastDirtyTimestamp();
//如果已存的微服务时间>要注册的(时间越大说明操作越新),用已存的覆盖要注册的
//即如果出现冲突的话拿最新的微服务实例
if (existingLastDirtyTimestamp > registrationLastDirtyTimestamp) {
registrant = existingLease.getHolder();
}
}
//没有拿到,一般都是进这里
else {
synchronized (lock) {
//期待发送心跳的客户端数量
if (this.expectedNumberOfClientsSendingRenews > 0) {
//要注册进来了,期待值+1
this.expectedNumberOfClientsSendingRenews = this.expectedNumberOfClientsSendingRenews + 1;
//更新客户端每分钟发送心跳数的阈值,这个方法较重要
updateRenewsPerMinThreshold();
}
}
}
//不管如何都new一个Lease
Lease<InstanceInfo> lease = new Lease(registrant, leaseDuration);
//如果if(gMap == null)都没有进,说明微服务组内已经有微服务了,直接put(id,instance)即可
((Map)gMap).put(registrant.getId(), lease);
//最近注册队列添加此微服务
recentRegisteredQueue.add(new Pair<Long, String>(
System.currentTimeMillis(),
registrant.getAppName() + "(" + registrant.getId() + ")"));
//标记微服务实例ADDED
registrant.setActionType(ActionType.ADDED);
//最近改变队列添加此微服务,此队列会保存近三分钟有改动的微服务,用于增量更新
recentlyChangedQueue.add(new RecentlyChangedItem(lease));
//设置最后更新的时间戳
registrant.setLastUpdatedTimestamp();
}
updateRenewsPerMinThreshold() protected void updateRenewsPerMinThreshold() {
this.numberOfRenewsPerMinThreshold = (int) (this.expectedNumberOfClientsSendingRenews
* (60.0 / serverConfig.getExpectedClientRenewalIntervalSeconds())
* serverConfig.getRenewalPercentThreshold());
}
expectedNumberOfClientsSendingRenews:期待发送心跳的客户端数量 ExpectedClientRenewalIntervalSeconds:期待客户端发送心跳的间隔秒数 RenewalPercentThreshold:续期的百分比阈值85% numberOfRenewsPerMinThreshold:客户端每分钟发送心跳数的阈值,如果server在一分钟内没有收到这么多的心跳数就会触发自我保护机制,以后博客应该会说到 举个例子就明白了: 假设有10个客户端,发送心跳间隔为30s,那么一分钟如果全部正常的话server收到的心跳应该是20次, 如果server一分钟收到的心跳<20*85%,即17个触发自我保护机制 PeerAwareInstanceRegistrImpl.replicateToPeers 集群同步 private void replicateToPeers(Action action, String appName, String id,
InstanceInfo info ,
InstanceStatus newStatus, boolean isReplication) {
//没有同伴,或者是isReplication=true则不要集群同步
//isReplication=true是这种情况,比如有两台eureka-server server1和server2组成了server集群
//此时有一个eureka-client user要注册到集群中,首先user注册到server1上(isReplication=false),
//server1会把user注册进registry这个map中然后集群同步将user注册给server2,此时
//isReplication就是true,这样server2注册好user后发现isReplication是true就不会再做集群同步,
//即不会给server1发送注册user的指令,避免了无限注册
if (peerEurekaNodes == Collections.EMPTY_LIST || isReplication) {
return;
}
for (final PeerEurekaNode node : peerEurekaNodes.getPeerEurekaNodes()) {
//跳过自己
if (peerEurekaNodes.isThisMyUrl(node.getServiceUrl())) {
continue;
}
replicateInstanceActionsToPeers(action, appName, id, info, newStatus, node);
}
}
private void replicateInstanceActionsToPeers(Action action, String appName,
String id, InstanceInfo info, InstanceStatus newStatus,
PeerEurekaNode node) {
switch (action) {
case Cancel:
node.cancel(appName, id);
break;
case Heartbeat:
InstanceStatus overriddenStatus = overriddenInstanceStatusMap.get(id);
infoFromRegistry = getInstanceByAppAndId(appName, id, false);
node.heartbeat(appName, id, infoFromRegistry, overriddenStatus, false);
break;
case Register:
node.register(info);
break;
case StatusUpdate:
infoFromRegistry = getInstanceByAppAndId(appName, id, false);
node.statusUpdate(appName, id, newStatus, infoFromRegistry);
break;
case DeleteStatusOverride:
infoFromRegistry = getInstanceByAppAndId(appName, id, false);
node.deleteStatusOverride(appName, id, infoFromRegistry);
break;
}
}
从集群同步的代码大致可以推断出流程如上图 集群中有三个server,当user模块注册到集群中的时候会先注册到一台server上,然后该server会遍历其他集群节点把user注册上去
总结
1:@EnableEurekaServer注入一个Marker类,说明是一个注册中心 2:EurekaServerAutoConfiguration注入一个filter,来拦截jersey请求转发给resource 3:服务注册,就是把信息存到一个ConcurrentHashMap<name,Map<id,Lease>> 4:对于注册冲突拿最新的微服务实例 5:server每分钟内收到的心跳数低于理应收到的85%就会触发自我保护机制 6:Lease的renew bug, duration多加了一次,理应加一个expireTime表示过期时间 7:集群同步:先注册到一台server,然后遍历其他的集群的其他server节点调用register注册到其他server, isReplication=true代表此次注册来源于集群同步的注册,代表此次注册不要再进行集群同步,避免无限注册
|