前一段时间，在一次开组会的时候，给小组成员简单介绍了一下 Raft 协议。大概四年前读过 Raft 的论文，这次分享的时候，好多好多细节都忘了。所以，再次把 《In Search of an Understandable Consensus Algorithm》 这篇论文找出来，重读一遍，做个笔记和摘要，方便后续学习和复习。
Abstract Raft is a consensus algorithm for managing a replicated log. 开篇摘要就点出了 Raft 的特点： Raft 是一种管理复制日志的共识算法。
In order to enhance understandability, Raft separates the key elements of consensus, such as leader election, log replication, and safety, and it enforcesa stronger degree of coherency to reduce the number of states that must be considered. 为了增强可理解性，Raft 将共识分解成几个关键元素，例如 Leader 选举，日志复制，以及安全性等；同时，为了降低需要考虑的状态的数量，还强制实施了更强的一致性。
1. Introduction Consensus algorithms allow a collection of machines to work as a coherent group that can survive the failures of some of its members.
在文章 Spring AOP 处理流程概述 中，对 Spring AOP 有了一个整体认识。在文章 Spring AOP 源码分析：入门 中，对 Spring AOP 的相关入口做了分析。这篇文章就带大家看一看，Spring AOP 是如何获取通知的？
示例代码 在 如何阅读 Spring 源码？: 示例代码 中，已经给出了一个完整的 AOP 示例代码。为了节省篇幅，请直接参考那篇文章的示例代码，这里就不在赘述。
注册 Advice(通知/增强) 请根据 Spring AOP 源码分析：入门 中提到的关键方法入口处，打上断点，开始调试。
首先，需要明确一点的是：对于切面（使用 @Aspect 注解标注过的类）在 Spring 容器中，也是被统一f封装为 BeanDefinition 实例的，也需要通过一个方式，将其注册到 Spring 容器中。比如，就像 示例代码 那样，通过 ImportSelector 方式，使用类名，将其注册到容器中。这样，就可以利用 Spring 容器对 Bean 的 API 来统一处理了。
Advice(通知/增强)几乎是在意想不到的地方完成注册的：在第一次调用 AbstractAutoProxyCreator#postProcessBeforeInstantiation 方法时，通过 AspectJAwareAdvisorAutoProxyCreator#shouldSkip 方法，完成了切面的注册。下面，我们对这个过程抽丝剥茧，逐步分析。
先来看看 findCandidateAdvisors 方法：
AnnotationAwareAspectJAutoProxyCreator#findCandidateAdvisors @Override protected List<Advisor> findCandidateAdvisors() { // Add all the Spring advisors found according to superclass rules. //当使用注解方式配置AOP的时候并不是丢弃了对XML配置的支持 //在这里调用父类方法加载配置文件中的AOP声明 List<Advisor> advisors = super.findCandidateAdvisors(); // Build Advisors for all AspectJ aspects in the bean factory. if (this.aspectJAdvisorsBuilder != null) { advisors.addAll(this.aspectJAdvisorsBuilder.buildAspectJAdvisors()); } return advisors; }
循环依赖在编程中是一个常见问题（当然，这并不是最佳实践）。并且，Spring 如何解决循环依赖这个问题在面试中也经常见。下面，D瓜哥就从源码的层面深入剖析一下这个问题。
示例程序 先展示一下示例程序：
package com.diguage.truman.context; import org.apache.commons.logging.Log; import org.apache.commons.logging.LogFactory; import org.junit.jupiter.api.Test; import org.springframework.beans.factory.annotation.Autowired; import org.springframework.context.annotation.AnnotationConfigApplicationContext; import org.springframework.context.annotation.Configuration; import org.springframework.context.annotation.Import; import org.springframework.context.annotation.ImportSelector; import org.springframework.core.type.AnnotationMetadata; import org.springframework.stereotype.Component; /** * @author D瓜哥, https://www.diguage.com/ * @since 2020-05-24 13:02 */ public class CircularDependenceSingletonTest { public static final Log log = LogFactory.getLog(CircularDependenceSingletonTest.class); @Test public void test() { AnnotationConfigApplicationContext applicationContext = new AnnotationConfigApplicationContext(); applicationContext.register(Config.class); applicationContext.refresh(); log.info(applicationContext.getBean(A.class)); log.info(applicationContext.getBean(B.class)); log.info(applicationContext.getBean(C.class)); log.info("-A--------"); A a = applicationContext.getBean(A.class); log.info(a); log.info(a.b); log.info("-B--------"); B b = applicationContext.getBean(B.class); log.info(b); log.info(b.c); log.info("-C--------"); C c = applicationContext.getBean(C.class); log.info(c); log.info(c.a); } @Configuration @Import(AbcImportSelector.class) public static class Config { } public static class AbcImportSelector implements ImportSelector { @Override public String[] selectImports(AnnotationMetadata importingClassMetadata) { return new String[]{ A.class.getName(), B.class.getName(), C.class.getName()}; } } @Component public static class A { @Autowired B b; } @Component public static class B { @Autowired C c; } @Component public static class C { @Autowired A a; } } 上述示例代码中的循环依赖情况如下：
图 1. 循环依赖 源码剖析 三级缓存 D瓜哥在 深入剖析 Spring 核心数据结构：BeanFactory 中，概要性地对 BeanFactory 的属性做了一一说明。 而其中的“三级缓存”属性，则是解决循环依赖问题的关键所在：
Map<String, Object> singletonObjects = new ConcurrentHashMap<>(256)：Bean 名称到单例 Bean 的映射，用于存放完全初始化好的 Bean。可以理解成，这就是所谓的容器。这是一级缓存。
Map<String, Object> earlySingletonObjects = new HashMap<>(16)：Bean 到“未成熟”单例 Bean 的映射。该 Bean 对象只是被创建出来，但是还没有注入依赖。在容器解决循环依赖时，用于存储中间状态。这是二级缓存。
Map<String, ObjectFactory<?>> singletonFactories = new HashMap<>(16)：Bean 名称到 Bean 的 ObjectFactory 对象的映射，存放 Bean 工厂对象。在容器解决循环依赖时，用于存储中间状态。这是三级缓存。
Bean 的获取过程就类似计算机缓存的作用过程：先从一级获取，失败再从二级、三级里面获取。在 org.springframework.beans.factory.support.DefaultSingletonBeanRegistry#getSingleton(java.lang.String, boolean) 方法中，可以明确看到整个过程：
org.springframework.beans.factory.support.DefaultSingletonBeanRegistry#getSingleton(beanName, allowEarlyReference) /** * Return the (raw) singleton object registered under the given name. * <p>Checks already instantiated singletons and also allows for an early * reference to a currently created singleton (resolving a circular reference). * @param beanName the name of the bean to look for * @param allowEarlyReference whether early references should be created or not * @return the registered singleton object, or {@code null} if none found */ @Nullable protected Object getSingleton(String beanName, boolean allowEarlyReference) { Object singletonObject = this.singletonObjects.get(beanName); if (singletonObject == null && isSingletonCurrentlyInCreation(beanName)) { synchronized (this.singletonObjects) { singletonObject = this.earlySingletonObjects.get(beanName); if (singletonObject == null && allowEarlyReference) { ObjectFactory<?> singletonFactory = this.singletonFactories.get(beanName); if (singletonFactory != null) { singletonObject = singletonFactory.getObject(); this.earlySingletonObjects.put(beanName, singletonObject); this.singletonFactories.remove(beanName); } } } } return singletonObject; }
在上一篇文章 分布式锁之 Apache Curator InterProcessMutex 中介绍了基于 ZooKeeper 实现的互斥锁。除此之外，还可以实现读写锁。这篇文章就来简要介绍一下 InterProcessReadWriteLock 的实现原理。
老规矩，先看看类的注释：
/** * <p> * A re-entrant read/write mutex that works across JVMs. Uses Zookeeper to hold the lock. All processes * in all JVMs that use the same lock path will achieve an inter-process critical section. Further, this mutex is * "fair" - each user will get the mutex in the order requested (from ZK's point of view). * </p> * * <p> * A read write lock maintains a pair of associated locks, one for read-only operations and one * for writing. The read lock may be held simultaneously by multiple reader processes, so long as * there are no writers. The write lock is exclusive. * </p> * * <p> * <b>Reentrancy</b><br> * This lock allows both readers and writers to reacquire read or write locks in the style of a * re-entrant lock. Non-re-entrant readers are not allowed until all write locks held by the * writing thread/process have been released. Additionally, a writer can acquire the read lock, but not * vice-versa. If a reader tries to acquire the write lock it will never succeed.<br><br> * * <b>Lock downgrading</b><br> * Re-entrancy also allows downgrading from the write lock to a read lock, by acquiring the write * lock, then the read lock and then releasing the write lock. However, upgrading from a read * lock to the write lock is not possible. * </p> */ public class InterProcessReadWriteLock {
