如何用源码分析canal的deployer模块

发布时间:2021-10-27 16:24:32 作者:柒染
来源:亿速云 阅读:166

这期内容当中小编将会给大家带来有关如何用源码分析canal的deployer模块,文章内容丰富且以专业的角度为大家分析和叙述,阅读完这篇文章希望大家可以有所收获。

  1. 获取canal.properties配置文件

  2. 如果canal.properties配置文件中属性root.admin.manager有值,那么构造PlainCanalConfigClient,调用PlainCanalConfigClient的findServer获取PlainCanal,调用PlainCanal的getProperties方法获取properties

  3. 通过properties构造 CanalStarter并调用其start方法

CanalStarter是启动类

public synchronized void start() throws Throwable {
        //首先根据canal.serverMode构造CanalMQProducer,如果是kafka,构造的是CanalKafkaProducer
        String serverMode = CanalController.getProperty(properties, CanalConstants.CANAL_SERVER_MODE);
        if (serverMode.equalsIgnoreCase("kafka")) {
            canalMQProducer = new CanalKafkaProducer();
        } else if (serverMode.equalsIgnoreCase("rocketmq")) {
            canalMQProducer = new CanalRocketMQProducer();
        }

        if (canalMQProducer != null) {
            // disable netty
            System.setProperty(CanalConstants.CANAL_WITHOUT_NETTY, "true");
            // 设置为raw避免ByteString->Entry的二次解析
            System.setProperty("canal.instance.memory.rawEntry", "false");
        }
        //接下来构造CanalController并调用其start方法
        logger.info("## start the canal server.");
        controller = new CanalController(properties);
        controller.start();
        logger.info("## the canal server is running now ......");
        ...
        //构造CanalMQStarter并调用其start方法,同时设置为CanalController的属性
        if (canalMQProducer != null) {
            canalMQStarter = new CanalMQStarter(canalMQProducer);
            MQProperties mqProperties = buildMQProperties(properties);
            String destinations = CanalController.getProperty(properties, CanalConstants.CANAL_DESTINATIONS);
            canalMQStarter.start(mqProperties, destinations);
            controller.setCanalMQStarter(canalMQStarter);
        }
        ...
        running = true;
    }
public CanalController(final Properties properties){
        // 初始化managerClients用于请求admin
        managerClients = MigrateMap.makeComputingMap(new Function<String, PlainCanalConfigClient>() {

            public PlainCanalConfigClient apply(String managerAddress) {
                return getManagerClient(managerAddress);
            }
        });

        // 初始化全局参数设置,包含了全局mode、lazy、managerAddress、springXml,初始化instanceGenerator用于创建instance,其根据InstanceConfig的mode值使用PlainCanalInstanceGenerator或者SpringCanalInstanceGenerator创建CanalInstance
        globalInstanceConfig = initGlobalConfig(properties);
        instanceConfigs = new MapMaker().makeMap();
        // 初始化instance config,包含了实例mode、lazy、managerAddress、springXml
        initInstanceConfig(properties);

        ...
        // 初始化CanalServerWithEmbedded,将instanceGenerator设置为CanalServerWithEmbedded的属性
        embededCanalServer = CanalServerWithEmbedded.instance();
        embededCanalServer.setCanalInstanceGenerator(instanceGenerator);// 设置自定义的instanceGenerator
        int metricsPort = Integer.valueOf(getProperty(properties, CanalConstants.CANAL_METRICS_PULL_PORT, "11112"));
        embededCanalServer.setMetricsPort(metricsPort);

        this.adminUser = getProperty(properties, CanalConstants.CANAL_ADMIN_USER);
        this.adminPasswd = getProperty(properties, CanalConstants.CANAL_ADMIN_PASSWD);
        embededCanalServer.setUser(getProperty(properties, CanalConstants.CANAL_USER));
        embededCanalServer.setPasswd(getProperty(properties, CanalConstants.CANAL_PASSWD));

        ...
        final String zkServers = getProperty(properties, CanalConstants.CANAL_ZKSERVERS);
        //初始化ZkClientx用于canal集群部署,创建/otteradmin/canal/destinations节点和/otteradmin/canal/cluster节点
        if (StringUtils.isNotEmpty(zkServers)) {
            zkclientx = ZkClientx.getZkClient(zkServers);
            // 初始化系统目录
            zkclientx.createPersistent(ZookeeperPathUtils.DESTINATION_ROOT_NODE, true);
            zkclientx.createPersistent(ZookeeperPathUtils.CANAL_CLUSTER_ROOT_NODE, true);
        }
        // 初始化ServerRunningMonitors的ServerRunningMonitor,用于启动、关闭实例
        final ServerRunningData serverData = new ServerRunningData(registerIp + ":" + port);
        ServerRunningMonitors.setServerData(serverData);
        ServerRunningMonitors.setRunningMonitors(MigrateMap.makeComputingMap(new Function<String, ServerRunningMonitor>() {
            ...
        }));

        // 初始化InstanceAction,用于启动和关闭实例
        autoScan = BooleanUtils.toBoolean(getProperty(properties, CanalConstants.CANAL_AUTO_SCAN));
        if (autoScan) {
            defaultAction = new InstanceAction() {
                ...
            };
            // 初始化instanceConfigMonitors,用于获取所有instanceConfig并启动所有instance
            instanceConfigMonitors = MigrateMap.makeComputingMap(new Function<InstanceMode, InstanceConfigMonitor>() {

                public InstanceConfigMonitor apply(InstanceMode mode) {
                    ...
                }
            });
        }
    }
public void start() {
        super.start();
        //启动定时任务,定时扫描所有instance
        executor.scheduleWithFixedDelay(new Runnable() {

            public void run() {
                try {
                    scan();
                    if (isFirst) {
                        isFirst = false;
                    }
                } catch (Throwable e) {
                    logger.error("scan failed", e);
                }
            }

        }, 0, scanIntervalInSecond, TimeUnit.SECONDS);
    }

private void scan() {
        //缓存了所有instance的配置,如果发现有新的instance则启动或者修改了instance则重启
        String instances = configClient.findInstances(null);
        final List<String> is = Lists.newArrayList(StringUtils.split(instances, ','));
        List<String> start = Lists.newArrayList();
        List<String> stop = Lists.newArrayList();
        List<String> restart = Lists.newArrayList();
        for (String instance : is) {
            if (!configs.containsKey(instance)) {
                PlainCanal newPlainCanal = configClient.findInstance(instance, null);
                if (newPlainCanal != null) {
                    configs.put(instance, newPlainCanal);
                    start.add(instance);
                }
            } else {
                PlainCanal plainCanal = configs.get(instance);
                PlainCanal newPlainCanal = configClient.findInstance(instance, plainCanal.getMd5());
                if (newPlainCanal != null) {
                    // 配置有变化
                    restart.add(instance);
                    configs.put(instance, newPlainCanal);
                }
            }
        }

        configs.forEach((instance, plainCanal) -> {
            if (!is.contains(instance)) {
                stop.add(instance);
            }
        });

        stop.forEach(instance -> {
            notifyStop(instance);
        });

        restart.forEach(instance -> {
            notifyReload(instance);
        });

        start.forEach(instance -> {
            notifyStart(instance);
        });

    }

private void notifyStart(String destination) {
        try {
            //启动instance调用InstanceAction启动实例,最后是调用ServerRunningMonitor启动实例
            defaultAction.start(destination);
            actions.put(destination, defaultAction);
            // 启动成功后记录配置文件信息
        } catch (Throwable e) {
            logger.error(String.format("scan add found[%s] but start failed", destination), e);
        }
    }
public ServerRunningMonitor(){
        // 创建父节点
        dataListener = new IZkDataListener() {

            public void handleDataChange(String dataPath, Object data) throws Exception {
                MDC.put("destination", destination);
                ServerRunningData runningData = JsonUtils.unmarshalFromByte((byte[]) data, ServerRunningData.class);
                if (!isMine(runningData.getAddress())) {
                    mutex.set(false);
                }

                if (!runningData.isActive() && isMine(runningData.getAddress())) { // 说明出现了主动释放的操作,并且本机之前是active
                    releaseRunning();// 彻底释放mainstem
                }

                activeData = (ServerRunningData) runningData;
            }

            public void handleDataDeleted(String dataPath) throws Exception {
                MDC.put("destination", destination);
                mutex.set(false);
                if (!release && activeData != null && isMine(activeData.getAddress())) {
                    // 如果上一次active的状态就是本机,则即时触发一下active抢占
                    initRunning();
                } else {
                    // 否则就是等待delayTime,避免因网络瞬端或者zk异常,导致出现频繁的切换操作
                    delayExector.schedule(new Runnable() {

                        public void run() {
                            initRunning();
                        }
                    }, delayTime, TimeUnit.SECONDS);
                }
            }

        };

    }

public synchronized void start() {
        super.start();
        try {
            //首先调用listener的processStart方法
            processStart();
            if (zkClient != null) {
                // 如果需要尽可能释放instance资源,不需要监听running节点,不然即使stop了这台机器,另一台机器立马会start
                // 监视/otteradmin/canal/destinations/{0}/running节点变化
                String path = ZookeeperPathUtils.getDestinationServerRunning(destination);
                zkClient.subscribeDataChanges(path, dataListener);
                
                initRunning();
            } else {
                processActiveEnter();// 没有zk,直接启动
            }
        } catch (Exception e) {
            logger.error("start failed", e);
            // 没有正常启动,重置一下状态,避免干扰下一次start
            stop();
        }

    }
private void processStart() {
        if (listener != null) {
            try {
                //processStart方法中创建/otteradmin/canal/destinations/{0}/cluster/{1}节点,0是实例名称,1是当前节点ip:port
                listener.processStart();
            } catch (Exception e) {
                logger.error("processStart failed", e);
            }
        }
    }

private void initRunning() {
        if (!isStart()) {
            return;
        }

        String path = ZookeeperPathUtils.getDestinationServerRunning(destination);
        // 序列化
        byte[] bytes = JsonUtils.marshalToByte(serverData);
        try {
            mutex.set(false);
            //尝试创建/otteradmin/canal/destinations/{0}/running节点
            zkClient.create(path, bytes, CreateMode.EPHEMERAL);
            activeData = serverData;
            //如果成功则调用listener的processEnter方法,processEnter方法中调用CanalServerWithEmbedded的start方法启动实例和CanalMQStarter的start方法启动实例
            processActiveEnter();// 触发一下事件
            mutex.set(true);
            release = false;
        } catch (ZkNodeExistsException e) {
            bytes = zkClient.readData(path, true);
            if (bytes == null) {// 如果不存在节点,立即尝试一次
                initRunning();
            } else {
                activeData = JsonUtils.unmarshalFromByte(bytes, ServerRunningData.class);
            }
        } catch (ZkNoNodeException e) {
            zkClient.createPersistent(ZookeeperPathUtils.getDestinationPath(destination), true); // 尝试创建父节点
            initRunning();
        }
    }
canal.register.ip =
canal.admin.manager = 127.0.0.1:8089
canal.admin.port = 11110
canal.admin.user = admin
canal.admin.passwd = 4ACFE3202A5FF5CF467898FC58AAB1D615029441
canal.admin.register.auto = true
canal.admin.register.cluster =

上述就是小编为大家分享的如何用源码分析canal的deployer模块了,如果刚好有类似的疑惑,不妨参照上述分析进行理解。如果想知道更多相关知识,欢迎关注亿速云行业资讯频道。

推荐阅读:
  1. 如何用jQuery 2.0.3源码分析Deferred
  2. nginx日志模块的源码分析

免责声明:本站发布的内容(图片、视频和文字)以原创、转载和分享为主,文章观点不代表本网站立场,如果涉及侵权请联系站长邮箱:is@yisu.com进行举报,并提供相关证据,一经查实,将立刻删除涉嫌侵权内容。

deployer canal

上一篇:在Linux文件系统中导航的使用技巧有哪些

下一篇:Mysql数据分组排名实现的示例分析

相关阅读

您好,登录后才能下订单哦!

密码登录
登录注册
其他方式登录
点击 登录注册 即表示同意《亿速云用户服务条款》