MQTT---HiveMQ源码详解(二)结构与启动

源博客地址:http://blog.csdn.net/pipinet123

---

MQTT交流群:221405150

---

目录结构

在官网中也有更详细的介绍，下面我只对目录结构做一个简单介绍即可，感兴趣的朋友可以参考官网文档.http://www.hivemq.com/docs/hivemq/latest/#installation

---

bin

包含hivemq.jar以及一些启动脚本

conf

包含config.xml、logback.xml以及plugin的配置文件
examples是一些示例组网场景的示例配置

data

metadata存放版本信息(加密过)
persistence存放着所有持久化信息的文件、以及备份文件。包含client_session_subscriptions、client_sessions、outgoing_message_flow、incomming_message_flow、publish_payloads、queued_messages、retained_messages等。

diagnostics

存放着诊断模式下诊断信息，包括系统信息、网络接口信息、jvm信息、插件信息等等。方便开发者排查问题。

license

存放hivemq授权license文件。

log

存放日志

plugins

第三方插件目录

---

启动

既然它是一个java程序,那么我们就从它的main方法开始我们的hivemq源码之路。

main

public class HiveMQServer {
    private static final Logger LOGGER = LoggerFactory.getLogger(HiveMQServer.class);
    private final NettyServer nettyServer;
    private final ClusterConfigurationService clusterConfigurationService;
    private final PluginBrokerCallbackHandler pluginBrokerCallbackHandler;
    private final PluginInformationStore pluginInformationStore;
    private final Provider clusterJoinerProvider;

    @Inject
    HiveMQServer(NettyServer nettyServer,
                 ClusterConfigurationService clusterConfigurationService,
                 PluginBrokerCallbackHandler pluginBrokerCallbackHandler,
                 PluginInformationStore pluginInformationStore,
                 Provider clusterJoinerProvider) {
        this.nettyServer = nettyServer;
        this.clusterConfigurationService = clusterConfigurationService;
        this.pluginBrokerCallbackHandler = pluginBrokerCallbackHandler;
        this.pluginInformationStore = pluginInformationStore;
        this.clusterJoinerProvider = clusterJoinerProvider;
    }

    public void start() throws InterruptedException, ExecutionException {
        //启动netty server
        this.nettyServer.start().sync();
        //通知OnBrokerStart事件
        fireOnBrokerStart();
        //加入cluster
        joinCluster();
        //启动对应承载在netty上的Listener，并打印出这些Listener启动结果信息。请参考Linstener配置请参考http://www.hivemq.com/docs/hivemq/latest/#configuration-chapter
        ListenableFuture> startFuture = this.nettyServer.startListeners();
        List startResults = startFuture.get();
        new ListenerStartResultLogger(startResults).log();
    }

    private void joinCluster() {
        //根据配置确定是否加入cluster
        if (!this.clusterConfigurationService.isEnabled()) {
            return;
        }
        try {
        //使用ClusterJoiner类进行连接jgroup，组成cluster。
            ClusterJoiner clusterJoiner = this.clusterJoinerProvider.get();
            ListenableFuture future = clusterJoiner.join();
            future.get();
        } catch (Exception e) {
            if (e.getCause() instanceof DuplicateOrInvalidLicenseException) {
                LOGGER.error("Found duplicate or invalid license file in the cluster. Shutting down HiveMQ");
            } else if (e.getCause() instanceof DifferentConfigurationException) {
                LOGGER.error("The configuration of this HiveMQ instance is different form the other instances in the cluster. Shutting down HiveMQ");
            } else {
                LOGGER.error("Could not join cluster. Shutting down HiveMQ.", e);
            }
            if (e.getCause() instanceof UnrecoverableException) {
                throw ((UnrecoverableException) e.getCause());
            }
            throw new UnrecoverableException(false);
        }
    }

    //通知对应plugin broker已经启动
    private void fireOnBrokerStart() {
        LOGGER.trace("Calling all OnBrokerStart Callbacks");
        printPluginInformations();
        this.pluginBrokerCallbackHandler.onStart();
    }

    public static void main(String[] args) throws InterruptedException, ExecutionException {
        LOGGER.info("Starting HiveMQ Server");
        long startTime = System.nanoTime();
        //初始化SystemInformation，可以通过环境变量来分别设置conf、plugins、log、license等目录。
        //请参考hivemq spi SystemInformation
        LOGGER.trace("Initializing HiveMQ home directory");
        HiveMQSystemInformation systemInformation = new HiveMQSystemInformation(true);
        //创建MetricRegistry
        //请参考开源框架Metrics
        LOGGER.trace("Creating MetricRegistry");
        MetricRegistry metricRegistry = new MetricRegistry();
        //增加统计Listener
        metricRegistry.addListener(new StatisticsListener());
        //初始化日志
        LOGGER.trace("Initializing Logging");
        LogConfigurator.init(systemInformation.getConfigFolder(), metricRegistry);
        //增加未处理异常拦截,并对其进行优雅处理
        LOGGER.trace("Initializing Exception handlers");
        RecoverableExceptionHandler.init();
        //初始化ConfigurationService,并读取conf/config.xml文件，加载用户配置
        //请参考hivemq spi ConfigurationService,
        LOGGER.trace("Initializing configuration");
        HiveMQConfigurationService hiveMQConfigurationService = HiveMQConfigurationServiceFactory.create(systemInformation);
        //创建Clusterid提供者。
        ClusterIdProducer clusterIdProducer = new ClusterIdProducer();
        if (hiveMQConfigurationService.clusterConfiguration().isEnabled()) {
            LOGGER.info("This node's cluster-ID is {}", clusterIdProducer.get());
        }
        //根据原有版本，判断是否需要做持久化数据的migration，如需要进行migration，因为可以配置每个数据的使用策略(file/memory)，所以每个数据分别进行migration
        LOGGER.trace("Checking for migrations");
        Map> neededMigrations = Migrations.getNeededMigrations(systemInformation);
        Injector injector = null;
        if (neededMigrations.size() > 0) {
            LOGGER.warn("HiveMQ has been updated, migrating persistent data to new version !");
            neededMigrations.keySet().forEach(type -> LOGGER.debug("{} needs to be migrated", type));
            //因为migration也是依赖guice来做容器，所以migration也会创建一个injector
            injector = Bootstrap.createInjector(systemInformation, hiveMQConfigurationService, clusterIdProducer);
            Migrations.start(injector, neededMigrations);
        }
        //升级完成,将升级的最新版本信息，持久化到文件中，以便下次启动进行判断
        Migrations.finish(systemInformation, hiveMQConfigurationService);
        //初始化guice
        LOGGER.trace("Initializing Guice");
        injector = Bootstrap.createInjector(systemInformation, metricRegistry, hiveMQConfigurationService, clusterIdProducer, injector);
        //从guice中获得HiveMQServer实例，并启动它
        HiveMQServer server = injector.getInstance(HiveMQServer.class);
        server.start();
        //对EXodus日志级别做修改
        LogConfigurator.addXodusLogModificator();
        LOGGER.info("Started HiveMQ in {}ms", TimeUnit.NANOSECONDS.toMillis(System.nanoTime() - startTime));
        //hivemq版本升级检查器，会连接hivemq官网判断是否有新版本升级。可以在配置文件中设置不检查
        UpdateChecker updateChecker = injector.getInstance(UpdateChecker.class);
        updateChecker.start();
    }

    //根据加载出来的所有plugin打印plugin信息
    //请参考hivemq spi @Information
    private void printPluginInformations() {
        Set pluginInformations = this.pluginInformationStore.getPluginInformations();
        pluginInformations.forEach(pluginInformation ->
                LOGGER.info("Loaded Plugin {} - v{}", pluginInformation.getName(), pluginInformation.getVersion())
        );
    }
}

Bootstrap & Guice Modules

它是采用Guice作为di框架，那么我们就从Bootstrap开始看它包含了哪些Module以及简单介绍下这些Module主要是注入哪些对应处理代码。

public class Bootstrap {
    private static final Logger LOGGER = LoggerFactory.getLogger(Bootstrap.class);

    public static Injector createInjector(SystemInformation systemInformation, MetricRegistry metricRegistry, HiveMQConfigurationService hiveMQConfigurationService, ClusterIdProducer clusterIdProducer, Injector injector) {
    //根据系统变量判断是否开启诊断模式
        if (!Boolean.parseBoolean(System.getProperty("diagnosticMode"))) {
            LOGGER.trace("Turning Guice stack traces off");
            System.setProperty("guice_include_stack_traces", "OFF");
        }
        //加载所有PluginModule
        //请参考hivemq spi PluginModule
        //后续会专门讲解plugin是如何加载的
        List pluginModules = new PluginBootstrap().create(systemInformation.getPluginFolder());
        ImmutableList.Builder builder = ImmutableList.builder();
        builder.add(
        //系统信息
                new SystemInformationModule(systemInformation),
           //注册cache的生命周期范围
                new ScopeModule(),
                //增加@PostConstruct、@PreDestroy注解处理
                new LifecycleModule(),
                //配置的Module
                new ConfigurationModule(hiveMQConfigurationService, clusterIdProducer),
                //netty所有handler、以及listenser等module
                new NettyModule(),
                //内部module
                new InternalModule(),
                //plugin callback module，主要处理plugin注册cabllback后回调
                new PluginCallbackModule(),
                //为方法增加cache的module
                new MethodCacheModule(),
                //持久化module
                new PersistenceModule(injector),
                //统计的module
                new MetricModule(metricRegistry),
                //流量监控module
                new TrafficShapingModule(),
                //cluster module
                new ClusterModule(),
                //plugin提供service的module
                new ServiceModule(pluginModules),
                //license的解析、验证、限制module
                new LicensingModule(),
                //更新hivemq程序的module
                new UpdateModule(),
                //诊断模式module
                new DiagnosticModule());
        builder.addAll(pluginModules);
        return Guice.createInjector(Stage.PRODUCTION, builder.build());
    }

//创建数据升级的Injector，这个较上面的module加载的少点而已。
    public static Injector createInjector(SystemInformation systemInformation,
                                          HiveMQConfigurationService hiveMQConfigurationService,
                                          ClusterIdProducer clusterIdProducer) {
        ImmutableList.Builder builder = ImmutableList.builder();
        builder.add(
                new SystemInformationModule(systemInformation),
                new ConfigurationModule(hiveMQConfigurationService, clusterIdProducer),
                new BridgeModule(),
                new ScopeModule(),
                new LifecycleModule());
        return Guice.createInjector(Stage.PRODUCTION, builder.build());
    }
}