flume可以监控并管理组件的运行状态，在组件关闭的时候可以自动拉起来，原理是通过启动一个计划任务线程池（monitorService ，线程的最大数量为30），运行监控线程（MonitorRunnable线程），每隔3s判断组件(包括Channel，SinkRunner)的状态是否符合要求(可用的状态由两种START和STOP)，根据不同的要求调用对应组件不同的方法，START会调用start方法，STOP会调用stop方法，如果想监控一个组件的状态，只需对这个组件调用supervise方法即可，如果想停止监控一个组件，只需对这个组件调用unsupervise方法即可，同时有一个线程每隔两小时移除已经不再监控（调用了unsupervise方法）的组件的检查任务。
这个功能主要是通过org.apache.flume.lifecycle.LifecycleSupervisor实现
在org.apache.flume.node.Application类的构造函数中会初始化LifecycleSupervisor类的对象：
public Application(List<LifecycleAware> components) {
this. components = components;
supervisor = new LifecycleSupervisor();
}flume进程启动时调用
org.apache.flume.node.Application.main--->org.apache.flume.node.Application.start
public synchronized void start() { //start方法会对每一个组件调用LifecycleSupervisor.supervise方法，参数为组件，AlwaysRestartPolicy和START状态（即期望的状态为START）
for(LifecycleAware component : components) {
supervisor.supervise(component,
new SupervisorPolicy.AlwaysRestartPolicy(), LifecycleState.START);
}
}分析LifecycleSupervisor类：
org.apache.flume.lifecycle.LifecycleSupervisor实现了LifecycleAware接口，本身也有一个生命周期的概念（提供start/stop等方法）
其定义了几个重要的内部类：
1.MonitorRunnable，实现了Runnable接口的线程类
1）3个属性ScheduledExecutorService monitorService，LifecycleAware lifecycleAware，Supervisoree supervisoree;
2）主要的run方法分析
public void run() {
long now = System.currentTimeMillis();
try {
if (supervisoree.status.firstSeen == null) {
logger.debug("first time seeing {}", lifecycleAware);
supervisoree.status.firstSeen = now; //第一次开始运行时，设置firstSeen为System.currentTimeMillis()
}
supervisoree.status.lastSeen = now; //设置lastSeen为now
synchronized (lifecycleAware) {
if (supervisoree.status.discard) { //如果Status的discard或者error的值为true，会直接退出
...
return;
} else if (supervisoree.status.error) {
...
return;
}
supervisoree.status.lastSeenState = lifecycleAware.getLifecycleState(); //设置lastSeenState的值
if (!lifecycleAware.getLifecycleState().equals(
supervisoree.status.desiredState)) { //如果获取的lifecycleAware对象状态不是想设置的desiredState状态
...
switch (supervisoree.status.desiredState) { //根据设置的desiredState状态调用lifecycleAware的不同方法,desiredState的值只有两种START和STOP
case START:
try {
lifecycleAware.start(); //状态为START时设置运行start方法
} catch (Throwable e) {
...
supervisoree.status.failures++; //start方法异常时failures的值加1
}
break;
case STOP:
try {
lifecycleAware.stop(); //状态为STOP时设置运行stop方法
} catch (Throwable e) {
...
supervisoree.status.failures++; //stop方法异常时failures的值加1
}
break;
default:
...
}
if (!supervisoree.policy.isValid(lifecycleAware, supervisoree.status)) {
//调用SupervisorPolicy的isValid方法，比如OnceOnlyPolicy 的isValid的方法会判断Status.failures 的值，如果为0则返回true,否则返回false
logger.error(
"Policy {} of {} has been violated - supervisor should exit!",
supervisoree.policy, lifecycleAware);
}
}
}
} catch(Throwable t) {
...
}
...
}2.Purger,实现了Runnable接口的线程类
run方法：
public void run() {
if(needToPurge){ //如果needToPurge设置为true
monitorService.purge(); //ScheduledThreadPoolExecutor.purge方法用于从工作队列中删除已经cancel的java.util.concurrent.Future对象（释放队列空间）
needToPurge = false ; //并设置needToPurge为false
}
}3.Status内部类定义了几个状态属性，代表了Supervisoree的状态
public Long firstSeen; public Long lastSeen; public LifecycleState lastSeenState; public LifecycleState desiredState; public int failures ; public boolean discard ; public volatile boolean error ;
4. SupervisorPolicy 是抽象类，定义了抽象方法isValid(LifecycleAware object, Status status)，包含两个扩展类AlwaysRestartPolicy 和OnceOnlyPolicy
AlwaysRestartPolicy 的isValid会一直返回true，OnceOnlyPolicy 的isValid的方法会判断Status.failures 的值，如果为0则返回true,否则返回false
5.Supervisoree包含SupervisorPolicy 和Status属性
主要的方法分析：
在构造方法中初始化几个重要的属性：
public LifecycleSupervisor() {
lifecycleState = LifecycleState.IDLE;
supervisedProcesses = new HashMap<LifecycleAware, Supervisoree>(); // supervisedProcesses 用于存放LifecycleAware和Supervisoree对象的键值对，代表已经管理的组件
monitorFutures = new HashMap<LifecycleAware, ScheduledFuture<?>>(); //monitorFutures 用于存放LifecycleAware对象和ScheduledFuture对象的键值对
monitorService = new ScheduledThreadPoolExecutor(10,
new ThreadFactoryBuilder().setNameFormat(
"lifecycleSupervisor-" + Thread.currentThread().getId() + "-%d")
.build()); // monitorService 用于调用Purger线程，定时移除线程池中已经cancel的task
monitorService.setMaximumPoolSize(20);
monitorService.setKeepAliveTime(30, TimeUnit.SECONDS);
purger = new Purger();
needToPurge = false; // 初始时为false，在有task cancel的时候设置为true
}
start方法用于启动检测线程池：
public synchronized void start() {
....
monitorService.scheduleWithFixedDelay( purger, 2, 2, TimeUnit. HOURS); //在两小时后每隔两小时运行一次Purger，释放线程池的工作队列
lifecycleState = LifecycleState. START; //设置状态为START
...
}
stop方法首先关闭线程池，然后关闭各个组件
1）线程池关闭
monitorService.shutdown();
2）各组件关闭
for ( final Entry<LifecycleAware, Supervisoree> entry : supervisedProcesses
.entrySet()) { //遍历supervisedProcesses中的各个组件
if (entry.getKey(). getLifecycleState().equals(LifecycleState.START)) { //如果组件的当前状态是START，则首先设置其需要变成的状态为STOP，并调用组件的stop方法
entry.getValue(). status. desiredState = LifecycleState.STOP;
entry.getKey().stop();
}
}supervise方法用于监控对应的组件，有3个参数LifecycleAware lifecycleAware, SupervisorPolicy policy, LifecycleState desiredState
public synchronized void supervise(LifecycleAware lifecycleAware,
SupervisorPolicy policy, LifecycleState desiredState) {
if(this. monitorService.isShutdown()
|| this.monitorService .isTerminated()
|| this.monitorService .isTerminating()){ //检测监控线程池是否正常
throw new FlumeException("Supervise called on " + lifecycleAware + " " +
"after shutdown has been initiated. " + lifecycleAware + " will not" +
" be started");
}
Preconditions.checkState(!supervisedProcesses .containsKey(lifecycleAware),
"Refusing to supervise " + lifecycleAware + " more than once" ); //检测是否已经管理
.....
Supervisoree process = new Supervisoree(); //初始化Supervisoree对象
process.status = new Status(); //并实例化Supervisoree对象的Status属性
process.policy = policy; //设置Supervisoree的属性
process.status.desiredState = desiredState;
process.status.error = false;
MonitorRunnable monitorRunnable = new MonitorRunnable(); //初始化一个MonitorRunnable 对象（线程），并设置对象的属性
monitorRunnable.lifecycleAware = lifecycleAware;
monitorRunnable.supervisoree = process;
monitorRunnable.monitorService = monitorService;
supervisedProcesses.put(lifecycleAware, process); //向supervisedProcesses中插入键值对，代表已经开始管理的组件
ScheduledFuture<?> future = monitorService.scheduleWithFixedDelay(
monitorRunnable, 0, 3, TimeUnit. SECONDS); // 设置计划任务线程池，每隔3s之后运行monitorRunnable
monitorFutures.put(lifecycleAware, future); // 向monitorFutures中插入键值对
}
unsupervise方法用于停止组件并从监控容器中去除：
synchronized (lifecycleAware) {
Supervisoree supervisoree = supervisedProcesses.get(lifecycleAware); //从已经管理的Supervisoree hashmap中获取Supervisoree对象
supervisoree.status.discard = true; //设置Supervisoree对象的Status属性的discard 值为discard
this.setDesiredState(lifecycleAware, LifecycleState.STOP);
//调用setDesiredState方法，设置Supervisoree对象的Status属性的desiredState 值为STOP（supervisoree.status.desiredState = desiredState）
logger.info("Stopping component: {}", lifecycleAware);
lifecycleAware.stop(); //调用组件的stop方法
}
supervisedProcesses.remove(lifecycleAware); //从supervisedProcesses hashmap中移除这个组件
monitorFutures.get(lifecycleAware).cancel(false);
//调用组件对应的ScheduledFuture的cancel方法取消任务（A Future represents the result of an asynchronous computation.
，cancel ：Attempts to cancel execution of this task.）
needToPurge = true; //设置needToPurge 的属性为true，这样就可以在purge中删除已经cancel的ScheduledFuture对象
monitorFutures.remove(lifecycleAware);本文出自 “菜光光的博客” 博客，请务必保留此出处http://caiguangguang.blog.51cto.com/1652935/1619527
原文地址：http://caiguangguang.blog.51cto.com/1652935/1619527
