1.什么是Akka Cluster？
Akka Cluster将多个JVM连接整合在一起，实现消息地址的透明化和统一化使用管理，集成一体化的消息驱动系统。最终目的是将一个大型程序分割成若干子程序，部署到很多JVM上去实现程序的分布式并行运算（单机也可以起很多节点构成集群）。更重要的是, Akka Cluster集群构建与Actor编程没有直接的联系，集群构建是在ActorSystem层面上，实现了Actor消息地址的透明化，无需考虑目标运行环节是否分布式，可以按照正常的Actor编程模式进行开发。 我们知道，分布式集群是由若干节点组成的，那么节点的发现及状态管理是分布式系统一个比较重要的任务。Akka Cluster中将节点的生命周期划分为：
- joining - 当尝试加入集群时的初始状态
- up - 加入集群后的正常状态
- leaving / exiting - 节点退出集群时的中间状态
- down - 集群无法感知某节点后，将其标记为down
- removed - 从集群中被删除，以后也无法再加入集群
其实当参数akka.cluster.allow-weakly-up-members启用时(默认是启用的)，还有个weakly up，它是用于集群出现分裂时，集群无法收敛，则leader无法将状态置为up的临时状态。这个后面再解释。 图中还有两个特殊的名词：
- fd* - 这个表示akka的错误检测机制
Faiulre Detector被触发后，将节点标记为unreachable - unreachable* -
unreachable不是一个真正的节点状态，更多的像是一个flag，用来描述集群无法与该节点进行通讯。当错误检测机制侦测到这个节点又能正常通讯时，会移除这个flag。
市面上大多数产品的分布式管理一般用的是注册中心机制，例如zk、consul或etcd。其实是节点把自己的信息注册到所使用的注册中心里，而master通过接受注册中心的通知得知新节点信息。显然本质上是一种master/slave的架构。这种架构有两个问题：
- master节点一般是单一的，一旦挂了影响就比较大（所以很多master都采用了HA机制），也就是所谓的系统单点故障；
- 通常节点的地址发现是要走master去获取的，当系统并发大时，master节点就可能成为性能瓶颈，即单点性能瓶颈。
Akka可能就是考虑这两点，采用了P2P的模式，这样任何一个节点都可以作为"master"，任何的节点都可以用来寻找其他节点地址。那它是怎么做到的呢？答案是Gossip协议和CRDT。这里不做过多解释，感兴趣的话可以自己去翻阅相关介绍
2.代码工程
实验目的
搭建一个简单akka custer集群
pom.xml
<!-- Akka Cluster dependency -->
<dependency>
<groupId>com.typesafe.akka</groupId>
<artifactId>akka-cluster-typed_2.13</artifactId>
<version>2.6.0</version>
</dependency>cluster
node1.conf
akka {
actor {
provider = "cluster"
}
remote {
artery {
canonical.hostname = "127.0.0.1"
canonical.port = 2551
}
}
cluster {
seed-nodes = [
"akka://ClusterSystem@127.0.0.1:2551",
"akka://ClusterSystem@127.0.0.1:2552"
]
}
}node2.conf
akka {
actor {
provider = "cluster"
}
remote {
artery {
canonical.hostname = "127.0.0.1"
canonical.port = 2552
}
}
cluster {
seed-nodes = [
"akka://ClusterSystem@127.0.0.1:2551",
"akka://ClusterSystem@127.0.0.1:2552"
]
}
}集群监听器
package com.et.akka.cluster;
import akka.actor.typed.Behavior;
import akka.actor.typed.javadsl.AbstractBehavior;
import akka.actor.typed.javadsl.ActorContext;
import akka.actor.typed.javadsl.Behaviors;
import akka.actor.typed.javadsl.Receive;
import akka.cluster.typed.Cluster;
import akka.cluster.typed.Subscribe;
import akka.cluster.ClusterEvent;
public class ClusterListener extends AbstractBehavior<ClusterEvent.ClusterDomainEvent> {
public ClusterListener(ActorContext<ClusterEvent.ClusterDomainEvent> context) {
super(context);
Cluster cluster = Cluster.get(context.getSystem());
cluster.subscriptions().tell(Subscribe.create(getContext().getSelf(), ClusterEvent.ClusterDomainEvent.class));
}
@Override
public Receive<ClusterEvent.ClusterDomainEvent> createReceive() {
return newReceiveBuilder()
.onMessage(ClusterEvent.MemberUp.class, this::onMemberUp)
.onMessage(ClusterEvent.MemberRemoved.class, this::onMemberRemoved)
.onAnyMessage(event -> {
System.out.println("Received cluster event: " + event);
return this;
})
.build();
}
private Behavior<ClusterEvent.ClusterDomainEvent> onMemberUp(ClusterEvent.MemberUp memberUp) {
System.out.println("Member is Up: " + memberUp.member());
return this;
}
private Behavior<ClusterEvent.ClusterDomainEvent> onMemberRemoved(ClusterEvent.MemberRemoved memberRemoved) {
System.out.println("Member is Removed: " + memberRemoved.member());
return this;
}
public static Behavior<ClusterEvent.ClusterDomainEvent> create() {
return Behaviors.setup(ClusterListener::new);
}
}启动集群
package com.et.akka.cluster;
import akka.actor.typed.ActorSystem;
import akka.cluster.ClusterEvent;
import com.typesafe.config.Config;
import com.typesafe.config.ConfigFactory;
import java.io.File;
public class ClusterApp {
public static void main(String[] args) {
Config configNode1 = ConfigFactory.parseFile(new File("D:/IdeaProjects/ETFramework/akka/src/main/resources/node1.conf"))
.withFallback(ConfigFactory.load());
ActorSystem<ClusterEvent.ClusterDomainEvent> systemNode1 = ActorSystem.create(ClusterListener.create(), "ClusterSystem", configNode1);
System.out.println("Node 1 started with config from node1.conf");
Config configNode2 = ConfigFactory.parseFile(new File("D:/IdeaProjects/ETFramework/akka/src/main/resources/node2.conf"))
.withFallback(ConfigFactory.load());
ActorSystem<ClusterEvent.ClusterDomainEvent> systemNode2 = ActorSystem.create(ClusterListener.create(), "ClusterSystem", configNode2);
System.out.println("Node 2 started with config from node2.conf");
}
}以上只是一些关键代码，所有代码请参见下面代码仓库
代码仓库
3.测试
启动集群（执行ClusterApp里面的main方法），查看日志可以看到2个节点都起来了
23:00:19.201 [ClusterSystem-akka.actor.default-dispatcher-6] INFO akka.cluster.Cluster - Cluster Node [akka://ClusterSystem@127.0.0.1:2552] - Welcome from [akka://ClusterSystem@127.0.0.1:2551]
Member is Up: Member(address = akka://ClusterSystem@127.0.0.1:2551, status = Up)
Received cluster event: MemberJoined(Member(address = akka://ClusterSystem@127.0.0.1:2552, status = Joining))
Received cluster event: LeaderChanged(Some(akka://ClusterSystem@127.0.0.1:2551))
Received cluster event: RoleLeaderChanged(dc-default,Some(akka://ClusterSystem@127.0.0.1:2551))
Received cluster event: SeenChanged(true,Set(akka://ClusterSystem@127.0.0.1:2551, akka://ClusterSystem@127.0.0.1:2552))
Received cluster event: ReachabilityChanged()
Received cluster event: SeenChanged(true,Set(akka://ClusterSystem@127.0.0.1:2551, akka://ClusterSystem@127.0.0.1:2552))
Received cluster event: ReachabilityChanged()
23:00:19.645 [ClusterSystem-akka.actor.default-dispatcher-5] INFO akka.cluster.Cluster - Cluster Node [akka://ClusterSystem@127.0.0.1:2551] - Leader is moving node [akka://ClusterSystem@127.0.0.1:2552] to [Up]
Member is Up: Member(address = akka://ClusterSystem@127.0.0.1:2552, status = Up)
Received cluster event: SeenChanged(false,Set(akka://ClusterSystem@127.0.0.1:2551))
Member is Up: Member(address = akka://ClusterSystem@127.0.0.1:2552, status = Up)
Received cluster event: ReachabilityChanged()
Received cluster event: SeenChanged(true,Set(akka://ClusterSystem@127.0.0.1:2551, akka://ClusterSystem@127.0.0.1:2552))
Received cluster event: ReachabilityChanged()