引言
RocketMq3.X的版本和Kafka一样是基于Zookeeper进行路由管理的,但是这意味着运维需要多部署一套Zookeeper集群,后来RocketMq选择去ZK最终出现了NameServ。NameServ作为RocketMq源码阅读的切入点非常不错,本文将会介绍Ver 4.9.4 版本的NameServ源码分析。
NameServer主要有两个功能,Broker管理和路由信息管理。
整个NameServ实际代码只有几百行,因为本身出现根本目的就是替代ZK,所以角色类似ZK。在下面的截图中,NamesrvStartup为启动类,NamesrvController为核心控制器,RouteInfoManager为路由信息表,整个NameServ基本上就是围绕这三个类做文章。
NameServe的类结构图如下:
源码分析
NameServ 启动
NameServ的启动步骤主要有下面几个点:
- 创建NameServ控制器,解析和创建重要配置,重要核心控制器创建并注入配置。
- NameServ核心控制器初始化,NettyServ服务等次重要相关组件创建和初始化。
- 启动定时任务,定期扫描过期Broker并且移除不活跃Broker,定期打印系统全部的KV配置。
- 注册JVM钩子函数优雅关闭资源(Netty和线程池),启动Netty。
- Netty服务启动
在了解代码细节之前,我们先画一个时序图了解NameServ的启动过程:
显然NameServ的整个启动基本上是在为Nettty做了一系列周边服务,Netty是网络通信的核心框架。
访问入口
整个NameServ的入口为org.apache.rocketmq.namesrv.NamesrvStartup#main0
,我们直接定位到相关代码。
public static NamesrvController main0(String[] args) {
try {
// 1. 构建核心控制器
NamesrvController controller = createNamesrvController(args);
// 2. 启动控制器
start(controller);
String tip = "The Name Server boot success. serializeType=" + RemotingCommand.getSerializeTypeConfigInThisServer();
log.info(tip);
System.out.printf("%s%n", tip);
return controller;
} catch (Throwable e) {
e.printStackTrace();
System.exit(-1);
}
return null;
}
构建核心控制器
NameServer一开始的工作是构建核心控制器,从整体上看主要做了下面几个操作:
- 调用Apach Commons CLI 命令行解析工具进行命令解析。
- 根据运行时参数生成commandLine命令行对象。
- 创建NamesrvConfig和NettyServerConfig对象,读取
-c
指定的配置文件路径解析配置文件。 - namesrvConfig 和 nettyServerConfig 对象进行初始化。
Apach Commons CLI 工具可以帮助开发者快速构建服务器启动命令参数,并且支持输出到列表。这里我们接着进入到 org.apache.rocketmq.namesrv.NamesrvStartup#createNamesrvController
一探究竟。进入之后发现代码还不少,所以我们拆成多个部分分析。
下面是完整的代码:
public static NamesrvController createNamesrvController(String[] args) throws IOException, JoranException {
System.setProperty(RemotingCommand.REMOTING_VERSION_KEY, Integer.toString(MQVersion.CURRENT_VERSION));
//PackageConflictDetect.detectFastjson();
Options options = ServerUtil.buildCommandlineOptions(new Options());
commandLine = ServerUtil.parseCmdLine("mqnamesrv", args, buildCommandlineOptions(options), new PosixParser());
if (null == commandLine) {
System.exit(-1);
return null;
}
final NamesrvConfig namesrvConfig = new NamesrvConfig();
final NettyServerConfig nettyServerConfig = new NettyServerConfig();
nettyServerConfig.setListenPort(9876);
if (commandLine.hasOption('c')) {
String file = commandLine.getOptionValue('c');
if (file != null) {
InputStream in = new BufferedInputStream(new FileInputStream(file));
properties = new Properties();
properties.load(in);
MixAll.properties2Object(properties, namesrvConfig);
MixAll.properties2Object(properties, nettyServerConfig);
namesrvConfig.setConfigStorePath(file);
System.out.printf("load config properties file OK, %s%n", file);
in.close();
}
}
if (commandLine.hasOption('p')) {
InternalLogger console = InternalLoggerFactory.getLogger(LoggerName.NAMESRV_CONSOLE_NAME);
MixAll.printObjectProperties(console, namesrvConfig);
MixAll.printObjectProperties(console, nettyServerConfig);
System.exit(0);
}
MixAll.properties2Object(ServerUtil.commandLine2Properties(commandLine), namesrvConfig);
if (null == namesrvConfig.getRocketmqHome()) {
System.out.printf("Please set the %s variable in your environment to match the location of the RocketMQ installation%n", MixAll.ROCKETMQ_HOME_ENV);
System.exit(-2);
}
LoggerContext lc = (LoggerContext) LoggerFactory.getILoggerFactory();
JoranConfigurator configurator = new JoranConfigurator();
configurator.setContext(lc);
lc.reset();
configurator.doConfigure(namesrvConfig.getRocketmqHome() + "/conf/logback_namesrv.xml");
log = InternalLoggerFactory.getLogger(LoggerName.NAMESRV_LOGGER_NAME);
MixAll.printObjectProperties(log, namesrvConfig);
MixAll.printObjectProperties(log, nettyServerConfig);
final NamesrvController controller = new NamesrvController(namesrvConfig, nettyServerConfig);
// remember all configs to prevent discard
controller.getConfiguration().registerConfig(properties);
return controller;
}
因为内容比较多,这里这里分段进行介绍,,首先是注册相关启动后命令:
System.setProperty(RemotingCommand.REMOTING_VERSION_KEY, Integer.toString(MQVersion.CURRENT_VERSION));
//PackageConflictDetect.detectFastjson();
// 创建命令行参数对象,这里定义了 -h 和 -n参数
Options options = ServerUtil.buildCommandlineOptions(new Options());
// 根据Options和运行时参数args生成命令行对象,buildCommandlineOptions定义了-c参数(Name server config properties file)和-p参数(Print all config item)
commandLine = ServerUtil.parseCmdLine("mqnamesrv", args, buildCommandlineOptions(options), new PosixParser());
if (null == commandLine) {
System.exit(-1);
return null;
}
ServerUtil.buildCommandlineOptions(new Options())
以及org.apache.rocketmq.namesrv.NamesrvStartup#buildCommandlineOptions
方法内部的逻辑:
// org.apache.rocketmq.srvutil.ServerUtil#buildCommandlineOptions
public static Options buildCommandlineOptions(final Options options) {
Option opt = new Option("h", "help", false, "Print help");
opt.setRequired(false);
options.addOption(opt);
opt =
new Option("n", "namesrvAddr", true,
"Name server address list, eg: '192.168.0.1:9876;192.168.0.2:9876'");
opt.setRequired(false);
options.addOption(opt);
return options;
}
// org.apache.rocketmq.namesrv.NamesrvStartup#buildCommandlineOptions
public static Options buildCommandlineOptions(final Options options) {
Option opt = new Option("c", "configFile", true, "Name server config properties file");
opt.setRequired(false);
options.addOption(opt);
opt = new Option("p", "printConfigItem", false, "Print all config items");
opt.setRequired(false);
options.addOption(opt);
return options;
}
从个人来看这个方法并不直观,并且复用性比较低,个人比较倾向于改成下面的方式:
public static Option buildCommandlineOption(String opt, String longOpt, boolean hasArg, String description, boolean required){
Option option = new Option(opt, longOpt, hasArg, description);
option.setRequired(required);
return option;
}
最后在本地个人把代码改造为下面的方式,虽然参数还需要优化,但是感觉直观了不少:
System.setProperty(RemotingCommand.REMOTING_VERSION_KEY, Integer.toString(MQVersion.CURRENT_VERSION));
//PackageConflictDetect.detectFastjson();
Options options = new Options();
// Modified to a more intuitive way of adding commands
options.addOption(ServerUtil.buildCommandlineOption("c", "configFile", true, "Name server config properties file", false));
options.addOption(ServerUtil.buildCommandlineOption("p", "printConfigItem", false, "Print all config items", false));
options.addOption(ServerUtil.buildCommandlineOption("h", "help", false, "Print help", false));
options.addOption(ServerUtil.buildCommandlineOption("n", "namesrvAddr", true,
"Name server address list, eg: '192.168.0.1:9876;192.168.0.2:9876'", false));
commandLine = ServerUtil.parseCmdLine("mqnamesrv", args, options, new PosixParser());
if (null == commandLine) {
System.exit(-1);
return null;}
如果觉得惹眼可以把这一段放到写好的方法里面,经过个人倒腾之后最终的代码如下:
System.setProperty(RemotingCommand.REMOTING_VERSION_KEY, Integer.toString(MQVersion.CURRENT_VERSION));
//PackageConflictDetect.detectFastjson();
Options options = buildCommandlineOptions(options);
commandLine = ServerUtil.parseCmdLine("mqnamesrv", args, options, new PosixParser());
if (null == commandLine) {
System.exit(-1);
return null;
}
//......
public static Options buildCommandlineOptions() {
Options options = new Options();
// Modified to a more intuitive way of adding commands
options.addOption(ServerUtil.buildCommandlineOption("c", "configFile", true, "Name server config properties file", false));
options.addOption(ServerUtil.buildCommandlineOption("p", "printConfigItem", false, "Print all config items", false));
options.addOption(ServerUtil.buildCommandlineOption("h", "help", false, "Print help", false));
options.addOption(ServerUtil.buildCommandlineOption("n", "namesrvAddr", true,
"Name server address list, eg: '192.168.0.1:9876;192.168.0.2:9876'", false));
return options;
}
相信读者对于Apach Commons CLI 命令行解析工具进行命令解析有了大致的了解。Apach的命令行解析工具帮助开发者根据运行时候的参数构建命令行对象,之后再通过 -c
的参数决定是否读取配置文件,解析配置文件之后填充到namesrvConfig和nettyServerConfig对象中。
解析命令之后是填充配置到对应的对象,填充配置文件的配置代码如下:
final NamesrvConfig namesrvConfig = new NamesrvConfig();
final NettyServerConfig nettyServerConfig = new NettyServerConfig();
nettyServerConfig.setListenPort(9876);
if (commandLine.hasOption('c')) {
String file = commandLine.getOptionValue('c');
if (file != null) {
InputStream in = new BufferedInputStream(new FileInputStream(file));
properties = new Properties();
properties.load(in);
MixAll.properties2Object(properties, namesrvConfig);
MixAll.properties2Object(properties, nettyServerConfig);
namesrvConfig.setConfigStorePath(file);
System.out.printf("load config properties file OK, %s%n", file);
in.close();
}
}
这一段算是createNamesrvController(String[] args)
核心代码之一,作用是先创建NettyServerConfig以及NamesrvConfig对象,然后利用commandLine命令行工具读取-c
指定的配置文件路径,这里用比较经典的缓冲流文件IO读取,之后生成Properties对象,这些代码基本都是JAVAEE基础,就不一一扣细节了。
当生成Properties对象完成之后,将namesrvConfig和nettyServerConfig对象进行初始化。接下来有一些不重要的代码,比如发现没有参数配置RocketMqHome会给出提示:
if (null == namesrvConfig.getRocketmqHome()) {
System.out.printf("Please set the %s variable in your environment to match the location of the RocketMQ installation%n", MixAll.ROCKETMQ_HOME_ENV);
System.exit(-2);
}
再比如会根据RocketMqHome的根路径下固定路径加载logback_namesrv.xml
日志配置文件,如果把日志重定向到自己其他磁盘路径,需要注意conf 这个层级文件夹以及日志配置文件一并拷贝。
LoggerContext lc = (LoggerContext) LoggerFactory.getILoggerFactory();
JoranConfigurator configurator = new JoranConfigurator();
configurator.setContext(lc);
lc.reset();
configurator.doConfigure(namesrvConfig.getRocketmqHome() + "/conf/logback_namesrv.xml");
之后便是重点操作创建NamesrvController核心控制器了,这里面把namesrvConfig和nettyServerConfig载入到核心控制器待后续初始化使用,代码如下:
final NamesrvController controller = new NamesrvController(namesrvConfig, nettyServerConfig);
// remember all configs to prevent discard
// 记住所有的配置以防止丢弃
controller.getConfiguration().registerConfig(properties);
上面的代码水到渠成地利用namesrvConfig和nettyServerConfig对象创建NamesrvController对象,然后在注册一遍properties防止丢失。
注意这里使用了JUC的 java.util.concurrent.locks.ReadWriteLock
读写锁进行操作
ReadWriteLock 是什么,可以参考廖老师的博客:使用ReadWriteLock - 廖雪峰的官方网站 (liaoxuefeng.com)
使用ReadWriteLock
可以提高读取效率:
ReadWriteLock
只允许一个线程写入;ReadWriteLock
允许多个线程在没有写入时同时读取;ReadWriteLock
适合读多写少的场景。
看完之后我们发现createNamesrvController(String[] args)
是非常重要的方法,内部的关键操作如下:
- 提供namesrvConfig和nettyServerConfig配置对象
- 创建NamesrvController核心控制器
创建完核心控制器之后紧接着便是启动控制器,这里有着次重要级别的初始化操作:
// 2. 启动控制器
start(controller);
初始化
创建核心控制器之后,紧接着是核心控制器的相关初始化动作,初始化的重要任务是下面几个:
- 初始化核心控制器,内部逻辑属于次重要级相关组件启动。
- 注册JVM钩子函数优雅关闭Netty和释放资源
- 核心控制器真正启动运行,实际上为触发Netty服务开启。
org.apache.rocketmq.namesrv.NamesrvStartup#start
初始化代码如下:
public static NamesrvController start(final NamesrvController controller) throws Exception {
if (null == controller) {
throw new IllegalArgumentException("NamesrvController is null");
}
// 对核心控制器进行初始化操作
boolean initResult = controller.initialize();
if (!initResult) {
controller.shutdown();
System.exit(-3);
}
// 注册一个钩子函数,JVM进程关闭时优雅地释放netty服务、线程池等资源
Runtime.getRuntime().addShutdownHook(new ShutdownHookThread(log, (Callable) () -> {
controller.shutdown();
return null; }));
// 核心控制器启动操作
controller.start();
return controller;
}
start()
的操作和创建核心控制器有点像,因为也是一个次重要级别的初始化操作。相关操作完成之后注册一个钩子函数优雅的释放Netty服务以及释放线程池的资源,最后对核心控制器进行启动操作。
我们继续深入核心控制器启动操作,org.apache.rocketmq.namesrv.NamesrvController#initialize
代码如下:
public boolean initialize() {
// 加载KV配置
this.kvConfigManager.load();
// 创建Netty网络服务对象
this.remotingServer = new NettyRemotingServer(this.nettyServerConfig, this.brokerHousekeepingService);
this.remotingExecutor =
Executors.newFixedThreadPool(nettyServerConfig.getServerWorkerThreads(), new ThreadFactoryImpl("RemotingExecutorThread_"));
this.registerProcessor();
// 创建定时任务--每个10s扫描一次Broker,并定时剔除不活跃的Broker
this.scheduledExecutorService.scheduleAtFixedRate(NamesrvController.this.routeInfoManager::scanNotActiveBroker, 5, 10, TimeUnit.SECONDS);
// 创建定时任务--每个10分钟打印一遍KV配置
this.scheduledExecutorService.scheduleAtFixedRate(NamesrvController.this.kvConfigManager::printAllPeriodically, 1, 10, TimeUnit.MINUTES);
// 省略SSL判断代码
return true;
}
这部分代码主要目的是对核心控制器进行启动前的一些初始化操作,包括下面一些内容:
- 根据上面方法初始化的NamesrvConfig的kvConfigPath(存储KV配置属性的路径)加载KV配置
创建两个定时任务:
- 每隔10s扫描一次Broker,并定时剔除不活跃的Broker
- 每隔10分钟打印一遍KV配置
这里的定时任务每次间隔10s扫描一次Broker,并定时剔除不活跃的Broker。
路由删除的逻辑放到后面进行介绍,这里暂时跳过
之后我们继续看核心控制器是如何启动的,方法入口为org.apache.rocketmq.namesrv.NamesrvController#start
:
public void start() throws Exception {
this.remotingServer.start();
if (this.fileWatchService != null) {
this.fileWatchService.start();
}
}
非常简单的,代码其实就是启动一下Netty服务罢了,因为RocketMq 底层通信是依赖Netty的,不过Netty的细节不在本文的讨论范围,这里就不过多介绍挖掘细节了。
至此整个路由启动的代码完成。
NameServ注册Broker
路由注册的时序图如下:
路由注册简单来说就是Broker注册到NameServ的过程,主要是通过心跳包实现的,那么Broker在代码中是如何存储的呢?我们根据上面的时序图最后一步就可以直接找到答案,就是在 RouteManager里面,里面维护了下面的信息:
private final Map> topicQueueTable;
private final Map brokerAddrTable;
private final Map > clusterAddrTable;
private final Map brokerLiveTable;
private final Map/* Filter Server */> filterServerTable;
private final Map> topicQueueMappingInfoTable;
和Spring 管理Bean差不多的套路,用的是万能的Map,上面定义的变量中比较重要的如下(和文章开头对应的一致):
- topicQueueTable:Topic消息队列路由信息,包括topic所在的broker名称,读队列数量,写队列数量,同步标记等信息,rocketmq根据topicQueueTable的信息进行负载均衡消息发送。
- brokerAddrTable:Broker节点信息,包括brokername,所在集群名称,还有主备节点信息。
- clusterAddrTable:Broker集群信息,存储了集群中所有的Brokername。
- brokerLiveTable:Broker状态信息,Nameserver每次收到Broker的心跳包就会更新该信息。
RocketMq在消息队列消费模式上使用的是发布订阅的模式设计,这在[[【RocketMq】RocketMq 扫盲]]中也有提到,这里不多赘述。
Broker中会存在一个Topic中有很多个Queue的情况,在默认的参数配置中RocketMq为每个新创建的Topic默认分配4个读队列和4个写队列,多个Broker还会组成集群,Broker还会定期向NameServ发送心跳包注册信息,NameServ则通过brokerLiveTable完成Broker节点状态的管理。
下面我们根据时序图一步步往下观察NameServ注册Broker的过程:
发送心跳包
上面我们分析了NameServ的启动代码,其实观察Broker的启动代码会发现有一定的相似之处,都是第一步构建一个控制器,然后start(),创建控制器这一部分内容不是重点这里跳过,我们接着看start()方法。
public static void main(String[] args) {
start(createBrokerController(args));
}
public static BrokerController start(BrokerController controller) {
try {
controller.start();
String tip = "The broker[" + controller.getBrokerConfig().getBrokerName() + ", "
+ controller.getBrokerAddr() + "] boot success. serializeType=" + RemotingCommand.getSerializeTypeConfigInThisServer();
if (null != controller.getBrokerConfig().getNamesrvAddr()) {
tip += " and name server is " + controller.getBrokerConfig().getNamesrvAddr();
}
log.info(tip);
System.out.printf("%s%n", tip);
return controller;
} catch (Throwable e) {
e.printStackTrace();
System.exit(-1);
}
return null;
}
controller.start();
是时序图的开始,下面是org.apache.rocketmq.broker.BrokerController#start:
的内部代码:
public void start() throws Exception {
// 此处省略相关依赖组件的start()过程
//.....
if (!messageStoreConfig.isEnableDLegerCommitLog()) {
startProcessorByHa(messageStoreConfig.getBrokerRole());
// 主从同步节点配置处理
handleSlaveSynchronize(messageStoreConfig.getBrokerRole());
// 初次启动强制发送心跳包
this.registerBrokerAll(true, false, true);
}
this.scheduledExecutorService.scheduleAtFixedRate(new Runnable() {
@Override
public void run() {
try {
BrokerController.this.registerBrokerAll(true, false, brokerConfig.isForceRegister());
} catch (Throwable e) {
log.error("registerBrokerAll Exception", e);
}
}
}, 1000 * 10, Math.max(10000, Math.min(brokerConfig.getRegisterNameServerPeriod(), 60000)), TimeUnit.MILLISECONDS);
if (this.brokerStatsManager != null) {
this.brokerStatsManager.start();
}
if (this.brokerFastFailure != null) {
this.brokerFastFailure.start();
}
}
registerBrokerAll 这个方法的参数可读性不太好,所以这里列举一下三个参数的顺序以及代码对应的参数数值:
- boolean checkOrderConfig(true)
- boolean oneway(false)
- boolean forceRegister(true)
搭配上参数之后就比较好懂了,也就是说加上配置校验以及强制执行一次注册动作,并且以非oneWay的方式发送一次心跳包。
下面我们顺利进入到 registerBrokerAll() 方法,方法内部首先创建topic包装类 ,然后会有一段比较有意思的代码,那就是如果没有读写权限会默认重新创建一个临时使用的topicConfigTable设置到Topic当中,之后是判断Broker此时是否需要执行发送心跳包。
但是我们回到上一级调用this.registerBrokerAll(true, false, true);
这里的参数传递就会发现,实际上forceRegister总是为true,也就是说基本上每个Broker第一次初始化必定需要传递心跳包的:
public synchronized void registerBrokerAll(final boolean checkOrderConfig, boolean oneway, boolean forceRegister) {
// 创建 TopicConfigSerializeWrapper,topic包装类
TopicConfigSerializeWrapper topicConfigWrapper = this.getTopicConfigManager().buildTopicConfigSerializeWrapper();
// 如果没有读写权限,此时会默认重新创建一个临时使用的topicConfigTable,作为Topic包装类的参数数值
// 个人认为这一步是防止空参数导致后面的方法出现异常,同时如果后续具备读写权限之后不需要重新创建直接使用
if (!PermName.isWriteable(this.getBrokerConfig().getBrokerPermission())
|| !PermName.isReadable(this.getBrokerConfig().getBrokerPermission())) {
// 这里初始化的值可以使用默认的Topic配置数量,比如加上topicConfigWrapper.getTopicConfigTable().values().size()
ConcurrentHashMap topicConfigTable = new ConcurrentHashMap<>(topicConfigWrapper.getTopicConfigTable().values().size());
for (TopicConfig topicConfig : topicConfigWrapper.getTopicConfigTable().values()) {
TopicConfig tmp =
new TopicConfig(topicConfig.getTopicName(), topicConfig.getReadQueueNums(), topicConfig.getWriteQueueNums(),
this.brokerConfig.getBrokerPermission());
topicConfigTable.put(topicConfig.getTopicName(), tmp);
}
topicConfigWrapper.setTopicConfigTable(topicConfigTable);
}
// 判断Broker是否需要发送心跳包
if (forceRegister || needRegister(this.brokerConfig.getBrokerClusterName(),
this.getBrokerAddr(),
this.brokerConfig.getBrokerName(),
this.brokerConfig.getBrokerId(),
this.brokerConfig.getRegisterBrokerTimeoutMills())) {
// 执行发送心跳包
doRegisterBrokerAll(checkOrderConfig, oneway, topicConfigWrapper);
}
}
下面我们接着定位到needRegister
方法进行解读,这里我们主要定位到org.apache.rocketmq.broker.out.BrokerOuterAPI#needRegister
方法,这里截取关键代码如下:
brokerOuterExecutor.execute(() -> {
try {
QueryDataVersionRequestHeader requestHeader = new QueryDataVersionRequestHeader();
requestHeader.setBrokerAddr(brokerAddr);
requestHeader.setBrokerId(brokerId);
requestHeader.setBrokerName(brokerName);
requestHeader.setClusterName(clusterName);
RemotingCommand request =
RemotingCommand.createRequestCommand(RequestCode.QUERY_DATA_VERSION, requestHeader);
request.setBody(topicConfigWrapper.getDataVersion().encode());
// 同步远程调用到路由中心
RemotingCommand response = remotingClient.invokeSync(namesrvAddr, request, timeoutMills);
DataVersion nameServerDataVersion = null;
Boolean changed = false;
// 省略代码:根据返回代码进行判断处理
//..
log.warn("Query data version from name server {} OK,changed {}, broker {},name server {}", namesrvAddr, changed, topicConfigWrapper.getDataVersion(), nameServerDataVersion == null ? "" : nameServerDataVersion);
} catch (Exception e) {
changedList.add(Boolean.TRUE);
log.error("Query data version from name server {} Exception, {}", namesrvAddr, e);
} finally {
countDownLatch.countDown();
}
});
这个代码不难理解,算是我们平常写HTTP调用的一个变体,我们可以通过RequestCode.QUERY_DATA_VERSION,查到NameServer的接受处理代码。
利用IDEA我们很快发现org.apache.rocketmq.namesrv.processor.DefaultRequestProcessor#processRequest
方法入口,之后进入到org.apache.rocketmq.namesrv.processor.DefaultRequestProcessor#queryBrokerTopicConfig
方法,然后这里看到对应代码如下:
public RemotingCommand queryBrokerTopicConfig(ChannelHandlerContext ctx,
RemotingCommand request) throws RemotingCommandException {
final RemotingCommand response = RemotingCommand.createResponseCommand(QueryDataVersionResponseHeader.class);
final QueryDataVersionResponseHeader responseHeader = (QueryDataVersionResponseHeader) response.readCustomHeader();
final QueryDataVersionRequestHeader requestHeader =
(QueryDataVersionRequestHeader) request.decodeCommandCustomHeader(QueryDataVersionRequestHeader.class);
DataVersion dataVersion = DataVersion.decode(request.getBody(), DataVersion.class);
// 内部处理:如果dataVersion为空或者当前dataVersion不等于brokerLiveTable存储的brokerLiveTable,Broker就需要发送心跳包
Boolean changed = this.namesrvController.getRouteInfoManager().isBrokerTopicConfigChanged(requestHeader.getBrokerAddr(), dataVersion);
if (!changed) {
// 更新Broker信息
this.namesrvController.getRouteInfoManager().updateBrokerInfoUpdateTimestamp(requestHeader.getBrokerAddr(), System.currentTimeMillis());
}
DataVersion nameSeverDataVersion = this.namesrvController.getRouteInfoManager().queryBrokerTopicConfig(requestHeader.getBrokerAddr());
response.setCode(ResponseCode.SUCCESS);
response.setRemark(null);
if (nameSeverDataVersion != null) {
response.setBody(nameSeverDataVersion.encode());
}
responseHeader.setChanged(changed);
return response;
}
我们进入到关键判断代码org.apache.rocketmq.namesrv.routeinfo.RouteInfoManager#isBrokerTopicConfigChanged
:
public boolean isBrokerTopicConfigChanged(final String brokerAddr, final DataVersion dataVersion) {
DataVersion prev = queryBrokerTopicConfig(brokerAddr);
// 如果dataVersion为空或者当前dataVersion不等于brokerLiveTable存储的brokerLiveTable,Broker就需要发送心跳包
return null == prev || !prev.equals(dataVersion);
}
public DataVersion queryBrokerTopicConfig(final String brokerAddr) {
BrokerLiveInfo prev = this.brokerLiveTable.get(brokerAddr);
if (prev != null) {
return prev.getDataVersion();
}
return null;
}
Broker是否需要发送心跳包由该Broker在路由中心org.apache.rocketmq.namesrv.routeinfo.BrokerLiveInfo#dataVersion
决定,如果dataVersion为空或者当前dataVersion不等于brokerLiveTable存储的brokerLiveTable,Broker就需要发送心跳包。
Nameserver处理心跳包
Nameserver的netty服务监听收到心跳包之后,会调用到路由中心以下方法进行处理,具体的方法入口为:org.apache.rocketmq.namesrv.routeinfo.RouteInfoManager#registerBroker
public RegisterBrokerResult registerBroker(
final String clusterName,
final String brokerAddr,
final String brokerName,
final long brokerId,
final String haServerAddr,
final TopicConfigSerializeWrapper topicConfigWrapper,
final List filterServerList,
final Channel channel) {
RegisterBrokerResult result = new RegisterBrokerResult();
try {
try {
this.lock.writeLock().lockInterruptibly();
// 获取集群下所有的Broker,并将当前Broker加入clusterAddrTable,由于brokerNames是Set结构,并不会重复
Set brokerNames = this.clusterAddrTable.computeIfAbsent(clusterName, k -> new HashSet<>());
brokerNames.add(brokerName);
boolean registerFirst = false;
// 获取Broker信息,如果是首次注册,那么新建一个BrokerData并加入brokerAddrTable
BrokerData brokerData = this.brokerAddrTable.get(brokerName);
if (null == brokerData) {
registerFirst = true;
brokerData = new BrokerData(clusterName, brokerName, new HashMap<>());
this.brokerAddrTable.put(brokerName, brokerData);
}
Map brokerAddrsMap = brokerData.getBrokerAddrs();
//Switch slave to master: first remove <1, IP:PORT> in namesrv, then add <0, IP:PORT>
//The same IP:PORT must only have one record in brokerAddrTable Iterator> it = brokerAddrsMap.entrySet().iterator();
//从库切换主库:首先删除namesrv中的<1, IP:PORT>,然后添加<0, IP:PORT>。
//同一个IP:端口在brokerAddrTable中只能有一条记录。
while (it.hasNext()) {
Entry item = it.next();
if (null != brokerAddr && brokerAddr.equals(item.getValue()) && brokerId != item.getKey()) {
log.debug("remove entry {} from brokerData", item);
it.remove();
}
}
// 里判断Broker是否是已经注册过
String oldAddr = brokerData.getBrokerAddrs().put(brokerId, brokerAddr);
registerFirst = registerFirst || (null == oldAddr);
// 如果是Broker是Master节点吗,并且Topic信息更新或者是首次注册,那么创建更新topic队列信息
if (null != topicConfigWrapper
&& MixAll.MASTER_ID == brokerId) {
if (this.isBrokerTopicConfigChanged(brokerAddr, topicConfigWrapper.getDataVersion())
|| registerFirst) {
ConcurrentMap tcTable =
topicConfigWrapper.getTopicConfigTable();
if (tcTable != null) {
for (Map.Entry entry : tcTable.entrySet()) {
this.createAndUpdateQueueData(brokerName, entry.getValue());
}
}
}
}
// 更新BrokerLiveInfo状态信息
BrokerLiveInfo prevBrokerLiveInfo = this.brokerLiveTable.put(brokerAddr,
new BrokerLiveInfo(
System.currentTimeMillis(),
topicConfigWrapper.getDataVersion(),
channel,
haServerAddr));
if (filterServerList != null) {
if (filterServerList.isEmpty()) {
this.filterServerTable.remove(brokerAddr);
} else {
this.filterServerTable.put(brokerAddr, filterServerList);
}
}
// 如果不是MASTER_ID,则返回结果返回masterAddr。
if (MixAll.MASTER_ID != brokerId) {
String masterAddr = brokerData.getBrokerAddrs().get(MixAll.MASTER_ID);
if (masterAddr != null) {
BrokerLiveInfo brokerLiveInfo = this.brokerLiveTable.get(masterAddr);
if (brokerLiveInfo != null) {
result.setHaServerAddr(brokerLiveInfo.getHaServerAddr());
result.setMasterAddr(masterAddr);
}
}
}
} finally {
this.lock.writeLock().unlock();
}
} catch (Exception e) {
log.error("registerBroker Exception", e);
}
return result;
}
上面的代码是Broker心跳包的最核心方法,它主要做了下面几件事:
RouteInfoManager路由信息的更新操作
- clusterAddrTable 更新;
- brokerAddrTable 更新;
- topicQueueTable 更新;
- brokerLiveTable 更新;
定期排除Broker
根据理论学习我们知道,NameServ在启动的时候会创建一个定时任务,定时剔除不活跃的Broker。这一点的源码在org.apache.rocketmq.namesrv.routeinfo.RouteInfoManager#scanNotActiveBroker
中可以找到答案。
此外在单元测试中就有关于这一项定期清理的测试,也是比较快的找到入口的办法:org.apache.rocketmq.namesrv.routeinfo.RouteInfoManagerBrokerRegisterTest#testScanNotActiveBroker
:
这个测试非常简单直观我们:
private static RouteInfoManager routeInfoManager;
public static String clusterName = "cluster";
public static String brokerPrefix = "broker";
public static String topicPrefix = "topic";
public static int brokerPerName = 3;
public static int brokerNameNumber = 3;
@Test
public void testScanNotActiveBroker() {
for (int j = 0; j < brokerNameNumber; j++) {
String brokerName = getBrokerName(brokerPrefix, j);
for (int i = 0; i < brokerPerName; i++) {
String brokerAddr = getBrokerAddr(clusterName, brokerName, i);
// set not active
routeInfoManager.updateBrokerInfoUpdateTimestamp(brokerAddr, 0);
assertEquals(1, routeInfoManager.scanNotActiveBroker());
}
}
}
在启动单元测试之前会先构建10个Broker节点注册进去,这里单元测试细心的使用了多个集群模拟生产环境:
private static RouteInfoManager routeInfoManager;
public static String clusterName = "cluster";
public static String brokerPrefix = "broker";
public static String topicPrefix = "topic";
public static int brokerPerName = 3;
public static int brokerNameNumber = 3;
@Before
public void setup() {
routeInfoManager = new RouteInfoManager();
cluster = registerCluster(routeInfoManager,
clusterName,
brokerPrefix,
brokerNameNumber,
brokerPerName,
topicPrefix,
10);
}
之后我们直接跑一边单元测试,在日志中单元测试为我们展示了详细的测试流程:
- 首先是构建broker注册,内部会塞入一些测试数据的Topic进行填充。
06:54:23.353 [main] INFO RocketmqNamesrv - new broker registered, cluster-broker-0:0 HAServer: cluster-broker-0:0
06:54:23.353 [main] INFO RocketmqNamesrv - new broker registered, cluster-broker-0:1 HAServer: cluster-broker-0:1
06:54:23.353 [main] INFO RocketmqNamesrv - new broker registered, cluster-broker-0:2 HAServer: cluster-broker-0:2
06:54:23.353 [main] INFO RocketmqNamesrv - cluster [cluster] brokerName [broker-1] master address change from null to cluster-broker-1:0
06:54:23.353 [main] INFO RocketmqNamesrv - new broker registered, cluster-broker-1:0 HAServer: cluster-broker-1:0
06:54:23.355 [main] INFO RocketmqNamesrv - new broker registered, cluster-broker-1:1 HAServer: cluster-broker-1:1
06:54:23.355 [main] INFO RocketmqNamesrv - new broker registered, cluster-broker-1:2 HAServer: cluster-broker-1:2
06:54:23.355 [main] INFO RocketmqNamesrv - cluster [cluster] brokerName [broker-2] master address change from null to cluster-broker-2:0
06:54:23.355 [main] INFO RocketmqNamesrv - new broker registered, cluster-broker-2:0 HAServer: cluster-broker-2:0
06:54:23.355 [main] INFO RocketmqNamesrv - new broker registered, cluster-broker-2:1 HAServer: cluster-broker-2:1
06:54:23.355 [main] INFO RocketmqNamesrv - new broker registered, cluster-broker-2:2 HAServer: cluster-broker-2:2
- 接着便是根据单元测试的代码进行遍历排除Broker节点,在循环的最后调用扫描检查不活跃Broker。这里为了验证直接设置lastUpdateTimestamp(最后更新时间)让Broker存活验证周期提前结束验证扫描效果。
06:55:34.483 [main] INFO RocketmqRemoting - closeChannel: close the connection to remote address[embedded] result: true
06:55:34.483 [main] WARN RocketmqNamesrv - The broker channel expired, cluster-broker-1:0 120000ms
06:55:34.483 [main] INFO RocketmqNamesrv - remove brokerAddr[0, cluster-broker-1:0] from brokerAddrTable, because channel destroyed
06:55:34.483 [main] INFO RocketmqRemoting - closeChannel: close the connection to remote address[embedded] result: true
06:55:34.483 [main] WARN RocketmqNamesrv - The broker channel expired, cluster-broker-1:1 120000ms
06:55:34.483 [main] INFO RocketmqNamesrv - remove brokerAddr[1, cluster-broker-1:1] from brokerAddrTable, because channel destroyed
06:55:34.483 [main] INFO RocketmqRemoting - closeChannel: close the connection to remote address[embedded] result: true
06:55:34.483 [main] WARN RocketmqNamesrv - The broker channel expired, cluster-broker-1:2 120000ms
06:55:34.484 [main] INFO RocketmqNamesrv - remove brokerAddr[2, cluster-broker-1:2] from brokerAddrTable, because channel destroyed
06:55:34.484 [main] INFO RocketmqNamesrv - remove brokerName[broker-1] from brokerAddrTable, because channel destroyed
06:55:34.484 [main] INFO RocketmqNamesrv - remove brokerName[broker-1], clusterName[cluster] from clusterAddrTable, because channel destroyed
以上便是单元测试的大致内容,我们接着看看具体的代码即可,这里还是用了迭代器模式进行遍历删除,又是一个经典的设计模式:
public int scanNotActiveBroker() {
int removeCount = 0;
Iterator> it = this.brokerLiveTable.entrySet().iterator();
while (it.hasNext()) {
Entry next = it.next();
long last = next.getValue().getLastUpdateTimestamp();
// BROKER_CHANNEL_EXPIRED_TIME = 1000 * 60 * 2 = 120秒,在单元测试中这里的last被设置为0所以必然超时
if ((last + BROKER_CHANNEL_EXPIRED_TIME) < System.currentTimeMillis()) {
RemotingUtil.closeChannel(next.getValue().getChannel());
it.remove();
log.warn("The broker channel expired, {} {}ms", next.getKey(), BROKER_CHANNEL_EXPIRED_TIME);
this.onChannelDestroy(next.getKey(), next.getValue().getChannel());
removeCount++;
}
}
return removeCount;
}
剔除Broker信息的逻辑比较简单,首先从BrokerLiveInfo获取状态信息,判断Broker的心跳时间是否已超过限定值(默认120秒),若超过之后就执行剔除操作。
写在最后
分析完了rocketmq自带的路由中心源码,其实我们自己实现一个路由中心貌似也不难。NameServ小而美的设计非常取巧,当然仅仅几百行代码确实还是存在比较多的不完美之处,很多方案需要开发人员自己编写业务代码兜底,但是有因为设计简单负责的任务,使用并且业务代码扩展性很强,维护成本低并且性能不错。
NameServ作为整个RocketMq的核心用法上简单的同时非常适合作为Rocketmq的切入点,个人在阅读代码中也会尝试修改代码查看效果,自己参与到源码编写和改造过程这会对代码编写者的思路更为清晰理解,也算是一个源码阅读的小技巧吧。