Dubbo分析之Registry层

前言
紧接上文Dubbo分析之Cluster层,本文继续分析dubbo的register层;此层封装服务地址的注册与发现,以服务URL为中心,扩展接口为RegistryFactory, Registry, RegistryService;

Registry接口
接口定义如下:

public interface Registry extends Node, RegistryService {
}

public interface RegistryService {

    void register(URL url);

    void unregister(URL url);
    
    void subscribe(URL url, NotifyListener listener);

    void unsubscribe(URL url, NotifyListener listener);

    List<URL> lookup(URL url);

}

主要提供了注册(register),注销(unregister),订阅(subscribe),退订(unsubscribe)等功能;dubbo提供了多种注册方式分别是:Multicast ,Zookeeper,Redis以及Simple方式;
Multicast:Multicast注册中心不需要启动任何中心节点,只要广播地址一样,就可以互相发现;
Zookeeper:Zookeeper是Apacahe Hadoop的子项目,是一个树型的目录服务,支持变更推送,适合作为Dubbo服务的注册中心,工业强度较高,可用于生产环境,并推荐使用;
Redis:基于Redis实现的注册中心,使用 Redis的Publish/Subscribe事件通知数据变更;
Simple:Simple注册中心本身就是一个普通的Dubbo服务,可以减少第三方依赖,使整体通讯方式一致;
后面重点介绍官方推荐的Zookeeper注册方式;具体的Register是在RegistryFactory中生成的,具体看一下接口定义;

RegistryFactory接口
接口定义如下:

@SPI("dubbo")
public interface RegistryFactory {

    @Adaptive({"protocol"})
    Registry getRegistry(URL url);

}

RegistryFactory提供了SPI扩展,默认使用dubbo,具体有哪些扩展可以查看META-INF/dubbo/internal/com.alibaba.dubbo.registry.RegistryFactory:

dubbo=com.alibaba.dubbo.registry.dubbo.DubboRegistryFactory
multicast=com.alibaba.dubbo.registry.multicast.MulticastRegistryFactory
zookeeper=com.alibaba.dubbo.registry.zookeeper.ZookeeperRegistryFactory
redis=com.alibaba.dubbo.registry.redis.RedisRegistryFactory

已推荐的Zookeeper为实例,查看ZookeeperRegistryFactory,提供了createRegistry方法:

 private ZookeeperTransporter zookeeperTransporter;
 
 public Registry createRegistry(URL url) {
        return new ZookeeperRegistry(url, zookeeperTransporter);
 }

实例化ZookeeperRegistry,两个参数分别是url和zookeeperTransporter,zookeeperTransporter是操作Zookeeper的客户端组件包括:zkclient和curator两种方式

@SPI("curator")
public interface ZookeeperTransporter {

    @Adaptive({Constants.CLIENT_KEY, Constants.TRANSPORTER_KEY})
    ZookeeperClient connect(URL url);

}

ZookeeperTransporter同样提供了SPI扩展,默认使用curator方式;接下来重点看一下Zookeeper注册中心。

Zookeeper注册中心
1.整体设计流程
在dubbo的整体设计中,可以大致查看Registry层的大致流程,首先通过RegistryFactory实例化Registry,Registry可以接收RegistryProtocol传过来的注册(register)和订阅(subscribe)消息,然后Registry通过ZKClient来向Zookeeper指定的目录下写入url信息,如果是订阅消息Registry会通过NotifyListener来通知RegitryDirctory进行更新url,最后就是Cluster层通过路由,负载均衡选择具体的提供方;

2.Zookeeper注册中心流程
官方提供了dubbo在Zookeeper中心的流程图:
zookeeper
流程说明:
服务提供者启动时: 向/dubbo/com.foo.BarService/providers目录下写入自己的URL地址;
服务消费者启动时: 订阅/dubbo/com.foo.BarService/providers目录下的提供者URL地址;并向/dubbo/com.foo.BarService/consumers目录下写入自己的URL地址;
监控中心启动时: 订阅/dubbo/com.foo.BarService 目录下的所有提供者和消费者URL地址。
下面分别从注册(register),注销(unregister),订阅(subscribe),退订(unsubscribe)四个方面来分析

3.注册(register)
ZookeeperRegistry的父类FailbackRegistry中实现了register方法,FailbackRegistry从名字可以看出来具有:失败自动恢复,后台记录失败请求,定时重发功能;下面具体看一下register方法:

public void register(URL url) {
        super.register(url);
        failedRegistered.remove(url);
        failedUnregistered.remove(url);
        try {
            // Sending a registration request to the server side
            doRegister(url);
        } catch (Exception e) {
            Throwable t = e;

            // If the startup detection is opened, the Exception is thrown directly.
            boolean check = getUrl().getParameter(Constants.CHECK_KEY, true)
                    && url.getParameter(Constants.CHECK_KEY, true)
                    && !Constants.CONSUMER_PROTOCOL.equals(url.getProtocol());
            boolean skipFailback = t instanceof SkipFailbackWrapperException;
            if (check || skipFailback) {
                if (skipFailback) {
                    t = t.getCause();
                }
                throw new IllegalStateException("Failed to register " + url + " to registry " + getUrl().getAddress() + ", cause: " + t.getMessage(), t);
            } else {
                logger.error("Failed to register " + url + ", waiting for retry, cause: " + t.getMessage(), t);
            }

            // Record a failed registration request to a failed list, retry regularly
            failedRegistered.add(url);
        }
    }

后台记录了失败的请求,包括failedRegistered和failedUnregistered,注册的时候将里面存放的url删除,然后执行doRegister方法,此方式在ZookeeperRegistry中实现,主要是在Zookeeper指定的目录下写入url信息,如果失败会记录注册失败的url,等待自动恢复;doRegister相关代码如下:

 protected void doRegister(URL url) {
        try {
            zkClient.create(toUrlPath(url), url.getParameter(Constants.DYNAMIC_KEY, true));
        } catch (Throwable e) {
            throw new RpcException("Failed to register " + url + " to zookeeper " + getUrl() + ", cause: " + e.getMessage(), e);
        }
}

调用zkClient的create方法在Zookeeper上创建节点,默认创建临时节点,create方法在AbstractZookeeperClient中实现,具体源码如下:

 public void create(String path, boolean ephemeral) {
        if (!ephemeral) {
            if (checkExists(path)) {
                return;
            }
        }
        int i = path.lastIndexOf('/');
        if (i > 0) {
            create(path.substring(0, i), false);
        }
        if (ephemeral) {
            createEphemeral(path);
        } else {
            createPersistent(path);
        }
    }

path指定需要创建的目录,ephemeral指定是否是创建临时节点,并且提供了递归创建目录,除了叶子目录其他目录都是持久化的;可以发现不管是创建临时目录还是持久化目录,都没有指定目录的Data,所有使用的是默认值,也就是本地ip地址;实例中创建的目录如下:

/dubbo/com.dubboApi.DemoService/providers/dubbo%3A%2F%2F10.13.83.7%3A20880%2Fcom.dubboApi.DemoService%3Fanyhost%3Dtrue%26application%3Dhello-world-app%26dubbo%3D2.0.2%26generic%3Dfalse%26interface%3Dcom.dubboApi.DemoService%26methods%3DsyncSayHello%2CsayHello%2CasyncSayHello%26pid%3D13252%26serialization%3Dprotobuf%26side%3Dprovider%26timestamp%3D1545297239027

dubbo是一个根节点,然后是service名称,providers是固定的一个类型,如果是消费端这里就是consumers,最后就是一个临时节点;使用临时节点的目的就是提供者出现断电等异常停机时,注册中心能自动删除提供者信息;可以通过如下方法查询当前的目录节点信息:

public class CuratorTest {

	static String path = "/dubbo";
	static CuratorFramework client = CuratorFrameworkFactory.builder().connectString("127.0.0.1:2181")
			.sessionTimeoutMs(5000).retryPolicy(new ExponentialBackoffRetry(1000, 3)).build();

	public static void main(String[] args) throws Exception {
		client.start();
		List<String> paths = listChildren(path);
		for (String path : paths) {
			Stat stat = new Stat();
			System.err.println(
					"path:" + path + ",value:" + new String(client.getData().storingStatIn(stat).forPath(path)));
		}
	}

	private static List<String> listChildren(String path) throws Exception {
		List<String> pathList = new ArrayList<String>();
		pathList.add(path);
		List<String> list = client.getChildren().forPath(path);
		if (list != null && list.size() > 0) {
			for (String cPath : list) {
				String temp = "";
				if ("/".equals(path)) {
					temp = path + cPath;
				} else {
					temp = path + "/" + cPath;
				}
				pathList.addAll(listChildren(temp));
			}
		}
		return pathList;
	}
}

递归遍历/dubbo目录下的所有子目录,同时将节点存储的数据都查询出来,结果如下:

path:/dubbo,value:10.13.83.7
path:/dubbo/com.dubboApi.DemoService,value:10.13.83.7
path:/dubbo/com.dubboApi.DemoService/configurators,value:10.13.83.7
path:/dubbo/com.dubboApi.DemoService/providers,value:10.13.83.7
path:/dubbo/com.dubboApi.DemoService/providers/dubbo%3A%2F%2F10.13.83.7%3A20880%2Fcom.dubboApi.DemoService%3Fanyhost%3Dtrue%26application%3Dhello-world-app%26dubbo%3D2.0.2%26generic%3Dfalse%26interface%3Dcom.dubboApi.DemoService%26methods%3DsyncSayHello%2CsayHello%2CasyncSayHello%26pid%3D4712%26serialization%3Dprotobuf%26side%3Dprovider%26timestamp%3D1545358401966,value:10.13.83.7

除了最后一个节点是临时节点,其他都是持久化的;

4.注销(unregister)
同样在父类FailbackRegistry中实现了unregister方法,代码如下:

 public void unregister(URL url) {
        super.unregister(url);
        failedRegistered.remove(url);
        failedUnregistered.remove(url);
        try {
            // Sending a cancellation request to the server side
            doUnregister(url);
        } catch (Exception e) {
            Throwable t = e;

            // If the startup detection is opened, the Exception is thrown directly.
            boolean check = getUrl().getParameter(Constants.CHECK_KEY, true)
                    && url.getParameter(Constants.CHECK_KEY, true)
                    && !Constants.CONSUMER_PROTOCOL.equals(url.getProtocol());
            boolean skipFailback = t instanceof SkipFailbackWrapperException;
            if (check || skipFailback) {
                if (skipFailback) {
                    t = t.getCause();
                }
                throw new IllegalStateException("Failed to unregister " + url + " to registry " + getUrl().getAddress() + ", cause: " + t.getMessage(), t);
            } else {
                logger.error("Failed to uregister " + url + ", waiting for retry, cause: " + t.getMessage(), t);
            }

            // Record a failed registration request to a failed list, retry regularly
            failedUnregistered.add(url);
        }
    }

注销时同样删除了failedRegistered和failedUnregistered存放的url,然后调用doUnregister,删除Zookeeper中的目录节点,失败的情况下会存储在failedUnregistered中,等待重试;

    protected void doUnregister(URL url) {
        try {
            zkClient.delete(toUrlPath(url));
        } catch (Throwable e) {
            throw new RpcException("Failed to unregister " + url + " to zookeeper " + getUrl() + ", cause: " + e.getMessage(), e);
        }
    }
	
	//CuratorZookeeperClient删除操作
	public void delete(String path) {
        try {
            client.delete().forPath(path);
        } catch (NoNodeException e) {
        } catch (Exception e) {
            throw new IllegalStateException(e.getMessage(), e);
        }
    }

直接使用CuratorZookeeperClient中的delete方法删除临时节点;

5.订阅(subscribe)
服务消费者启动时,会先向Zookeeper注册消费者节点信息,然后订阅…/providers目录下提供者的URL地址;消费端也同样需要注册节点信息,是因为监控中心需要对服务端和消费端都进行监控;下面重点看一下订阅的相关代码,在父类FailbackRegistry中实现了subscribe方法:

 public void subscribe(URL url, NotifyListener listener) {
        super.subscribe(url, listener);
        removeFailedSubscribed(url, listener);
        try {
            // Sending a subscription request to the server side
            doSubscribe(url, listener);
        } catch (Exception e) {
            Throwable t = e;

            List<URL> urls = getCacheUrls(url);
            if (urls != null && !urls.isEmpty()) {
                notify(url, listener, urls);
                logger.error("Failed to subscribe " + url + ", Using cached list: " + urls + " from cache file: " + getUrl().getParameter(Constants.FILE_KEY, System.getProperty("user.home") + "/dubbo-registry-" + url.getHost() + ".cache") + ", cause: " + t.getMessage(), t);
            } else {
                // If the startup detection is opened, the Exception is thrown directly.
                boolean check = getUrl().getParameter(Constants.CHECK_KEY, true)
                        && url.getParameter(Constants.CHECK_KEY, true);
                boolean skipFailback = t instanceof SkipFailbackWrapperException;
                if (check || skipFailback) {
                    if (skipFailback) {
                        t = t.getCause();
                    }
                    throw new IllegalStateException("Failed to subscribe " + url + ", cause: " + t.getMessage(), t);
                } else {
                    logger.error("Failed to subscribe " + url + ", waiting for retry, cause: " + t.getMessage(), t);
                }
            }

            // Record a failed registration request to a failed list, retry regularly
            addFailedSubscribed(url, listener);
        }
    }

类似的格式,同样存储了失败了订阅url信息,重点看ZookeeperRegistry中的doSubscribe方法:

 private final ConcurrentMap<URL, ConcurrentMap<NotifyListener, ChildListener>> zkListeners = new ConcurrentHashMap<URL, ConcurrentMap<NotifyListener, ChildListener>>();

 protected void doSubscribe(final URL url, final NotifyListener listener) {
        try {
            if (Constants.ANY_VALUE.equals(url.getServiceInterface())) {
                String root = toRootPath();
                ConcurrentMap<NotifyListener, ChildListener> listeners = zkListeners.get(url);
                if (listeners == null) {
                    zkListeners.putIfAbsent(url, new ConcurrentHashMap<NotifyListener, ChildListener>());
                    listeners = zkListeners.get(url);
                }
                ChildListener zkListener = listeners.get(listener);
                if (zkListener == null) {
                    listeners.putIfAbsent(listener, new ChildListener() {
                        @Override
                        public void childChanged(String parentPath, List<String> currentChilds) {
                            for (String child : currentChilds) {
                                child = URL.decode(child);
                                if (!anyServices.contains(child)) {
                                    anyServices.add(child);
                                    subscribe(url.setPath(child).addParameters(Constants.INTERFACE_KEY, child,
                                            Constants.CHECK_KEY, String.valueOf(false)), listener);
                                }
                            }
                        }
                    });
                    zkListener = listeners.get(listener);
                }
                zkClient.create(root, false);
                List<String> services = zkClient.addChildListener(root, zkListener);
                if (services != null && !services.isEmpty()) {
                    for (String service : services) {
                        service = URL.decode(service);
                        anyServices.add(service);
                        subscribe(url.setPath(service).addParameters(Constants.INTERFACE_KEY, service,
                                Constants.CHECK_KEY, String.valueOf(false)), listener);
                    }
                }
            } else {
                List<URL> urls = new ArrayList<URL>();
                for (String path : toCategoriesPath(url)) {
                    ConcurrentMap<NotifyListener, ChildListener> listeners = zkListeners.get(url);
                    if (listeners == null) {
                        zkListeners.putIfAbsent(url, new ConcurrentHashMap<NotifyListener, ChildListener>());
                        listeners = zkListeners.get(url);
                    }
                    ChildListener zkListener = listeners.get(listener);
                    if (zkListener == null) {
                        listeners.putIfAbsent(listener, new ChildListener() {
                            @Override
                            public void childChanged(String parentPath, List<String> currentChilds) {
                                ZookeeperRegistry.this.notify(url, listener, toUrlsWithEmpty(url, parentPath, currentChilds));
                            }
                        });
                        zkListener = listeners.get(listener);
                    }
                    zkClient.create(path, false);
                    List<String> children = zkClient.addChildListener(path, zkListener);
                    if (children != null) {
                        urls.addAll(toUrlsWithEmpty(url, path, children));
                    }
                }
                notify(url, listener, urls);
            }
        } catch (Throwable e) {
            throw new RpcException("Failed to subscribe " + url + " to zookeeper " + getUrl() + ", cause: " + e.getMessage(), e);
        }
    }

在ZookeeperRegistry中定义了一个zkListeners变量,每个URL对应了一个map;map里面分别是NotifyListener和ChildListener的对应关系,消费端订阅时这里的NotifyListener其实就是RegistryDirectory,ChildListener是一个内部类,用来在监听的节点发生变更时,通知对应的消费端,具体的监听处理是在zkClient.addChildListener中实现的:

    public List<String> addChildListener(String path, final ChildListener listener) {
        ConcurrentMap<ChildListener, TargetChildListener> listeners = childListeners.get(path);
        if (listeners == null) {
            childListeners.putIfAbsent(path, new ConcurrentHashMap<ChildListener, TargetChildListener>());
            listeners = childListeners.get(path);
        }
        TargetChildListener targetListener = listeners.get(listener);
        if (targetListener == null) {
            listeners.putIfAbsent(listener, createTargetChildListener(path, listener));
            targetListener = listeners.get(listener);
        }
        return addTargetChildListener(path, targetListener);
    }
	
	public CuratorWatcher createTargetChildListener(String path, ChildListener listener) {
        return new CuratorWatcherImpl(listener);
    }
	
	public List<String> addTargetChildListener(String path, CuratorWatcher listener) {
		try {
			return client.getChildren().usingWatcher(listener).forPath(path);
		} catch (NoNodeException e) {
			return null;
		} catch (Exception e) {
			throw new IllegalStateException(e.getMessage(), e);
		}
    }
	
	private class CuratorWatcherImpl implements CuratorWatcher {

        private volatile ChildListener listener;

        public CuratorWatcherImpl(ChildListener listener) {
            this.listener = listener;
        }

        public void unwatch() {
            this.listener = null;
        }

        @Override
        public void process(WatchedEvent event) throws Exception {
            if (listener != null) {
                String path = event.getPath() == null ? "" : event.getPath();
                listener.childChanged(path,
                        StringUtils.isNotEmpty(path)
                                ? client.getChildren().usingWatcher(this).forPath(path)
                                : Collections.<String>emptyList());
            }
        }
    }

CuratorWatcherImpl实现了Zookeeper的监听接口CuratorWatcher,用来在节点有变更时通知对应的ChildListener,这样ChildListener就可以通知RegistryDirectory进行更新数据;

6.退订(unsubscribe)
在父类FailbackRegistry中实现了unsubscribe方法

public void unsubscribe(URL url, NotifyListener listener) {
        super.unsubscribe(url, listener);
        removeFailedSubscribed(url, listener);
        try {
            // Sending a canceling subscription request to the server side
            doUnsubscribe(url, listener);
        } catch (Exception e) {
            Throwable t = e;

            // If the startup detection is opened, the Exception is thrown directly.
            boolean check = getUrl().getParameter(Constants.CHECK_KEY, true)
                    && url.getParameter(Constants.CHECK_KEY, true);
            boolean skipFailback = t instanceof SkipFailbackWrapperException;
            if (check || skipFailback) {
                if (skipFailback) {
                    t = t.getCause();
                }
                throw new IllegalStateException("Failed to unsubscribe " + url + " to registry " + getUrl().getAddress() + ", cause: " + t.getMessage(), t);
            } else {
                logger.error("Failed to unsubscribe " + url + ", waiting for retry, cause: " + t.getMessage(), t);
            }

            // Record a failed registration request to a failed list, retry regularly
            Set<NotifyListener> listeners = failedUnsubscribed.get(url);
            if (listeners == null) {
                failedUnsubscribed.putIfAbsent(url, new ConcurrentHashSet<NotifyListener>());
                listeners = failedUnsubscribed.get(url);
            }
            listeners.add(listener);
        }
    }

同样使用failedUnsubscribed用来存储失败退订的url,具体看ZookeeperRegistry中的doUnsubscribe方法

protected void doUnsubscribe(URL url, NotifyListener listener) {
        ConcurrentMap<NotifyListener, ChildListener> listeners = zkListeners.get(url);
        if (listeners != null) {
            ChildListener zkListener = listeners.get(listener);
            if (zkListener != null) {
                if (Constants.ANY_VALUE.equals(url.getServiceInterface())) {
                    String root = toRootPath();
                    zkClient.removeChildListener(root, zkListener);
                } else {
                    for (String path : toCategoriesPath(url)) {
                        zkClient.removeChildListener(path, zkListener);
                    }
                }
            }
        }
    }

退订就比较简单了,只需要移除监听器就可以了;

7.失败重试
FailbackRegistry从名字可以看出来具有:失败自动恢复,后台记录失败请求,定时重发功能;在FailbackRegistry的构造器中启动了一个定时器:

 this.retryFuture = retryExecutor.scheduleWithFixedDelay(new Runnable() {
            @Override
            public void run() {
                // Check and connect to the registry
                try {
                    retry();
                } catch (Throwable t) { // Defensive fault tolerance
                    logger.error("Unexpected error occur at failed retry, cause: " + t.getMessage(), t);
                }
            }
        }, retryPeriod, retryPeriod, TimeUnit.MILLISECONDS);

实例化了一个间隔5秒执行一次重试的定时器,retry部分代码如下:

 protected void retry() {
        if (!failedRegistered.isEmpty()) {
            Set<URL> failed = new HashSet<URL>(failedRegistered);
            if (failed.size() > 0) {
                if (logger.isInfoEnabled()) {
                    logger.info("Retry register " + failed);
                }
                try {
                    for (URL url : failed) {
                        try {
                            doRegister(url);
                            failedRegistered.remove(url);
                        } catch (Throwable t) { // Ignore all the exceptions and wait for the next retry
                            logger.warn("Failed to retry register " + failed + ", waiting for again, cause: " + t.getMessage(), t);
                        }
                    }
                } catch (Throwable t) { // Ignore all the exceptions and wait for the next retry
                    logger.warn("Failed to retry register " + failed + ", waiting for again, cause: " + t.getMessage(), t);
                }
            }
        }
        ...省略...
}

定期检查是否存在失败的注册(register),注销(unregister),订阅(subscribe),退订(unsubscribe)URL,如果存在则重试;

总结
本文首先介绍了RegistryFactory, Registry, RegistryService几个核心接口,然后以Zookeeper为注册中心重点介绍了注册(register),注销(unregister),订阅(subscribe),退订(unsubscribe)方式。