分布式锁—6.Redisson的同步器组件

大纲

1.Redisson的分布式锁简单总结

2.Redisson的Semaphore简介

3.Redisson的Semaphore源码剖析

4.Redisson的CountDownLatch简介

5.Redisson的CountDownLatch源码剖析

1.Redisson的分布式锁简单总结

(1)可重入锁RedissonLock

(2)公平锁RedissonFairLock

(3)联锁MultiLock

(4)红锁RedLock

(5)读写锁之读锁RedissonReadLock和写锁RedissonWriteLock

Redisson分布式锁包括：可重入锁、公平锁、联锁、红锁、读写锁。

(1)可重入锁RedissonLock

非公平锁，最基础的分布式锁，最常用的锁。

(2)公平锁RedissonFairLock

各个客户端尝试获取锁时会排队，按照队列的顺序先后获取锁。

(3)联锁MultiLock

可以一次性加多把锁，从而实现一次性锁多个资源。

(4)红锁RedLock

RedLock相当于一把锁。虽然利用了MultiLock包裹了多个小锁，但这些小锁并不对应多个资源，而是每个小锁的key对应一个Redis实例。只要大多数的Redis实例加锁成功，就可以认为RedLock加锁成功。RedLock的健壮性要比其他普通锁要好。

但是RedLock也有一些场景无法保证正确性，当然RedLock只要求部署主库。比如客户端A尝试向5个Master实例加锁，但仅仅在3个Maste中加锁成功。不幸的是此时3个Master中有1个Master突然宕机了，而且锁key还没同步到该宕机Master的Slave上，此时Salve切换为Master。于是在这5个Master中，由于其中有一个是新切换过来的Master，所以只有2个Master是有客户端A加锁的数据，另外3个Master是没有锁的。但继续不幸的是，此时客户端B来加锁，那么客户端B就很有可能成功在没有锁数据的3个Master上加到锁，从而满足了过半数加锁的要求，最后也完成了加锁，依然发生重复加锁。

(5)读写锁之读锁RedissonReadLock和写锁RedissonWriteLock

不同客户端线程的四种加锁情况：

情况一：先加读锁再加读锁，不互斥

情况二：先加读锁再加写锁，互斥

情况三：先加写锁再加读锁，互斥

情况四：先加写锁再加写锁，互斥

同一个客户端线程的四种加锁情况：

情况一：先加读锁再加读锁，不互斥

情况二：先加读锁再加写锁，互斥

情况三：先加写锁再加读锁，不互斥

情况四：先加写锁再加写锁，不互斥

2.Redisson的Semaphore简介

(1)Redisson的Semaphore原理图

Semaphore也是Redisson支持的一种同步组件。Semaphore作为一个锁机制，可以允许多个线程同时获取一把锁。任何一个线程释放锁之后，其他等待的线程就可以尝试继续获取锁。

(2)Redisson的Semaphore使用演示

public class RedissonDemo {
    public static void main(String[] args) throws Exception {
        //连接3主3从的Redis CLuster
        Config config = new Config();
        ...
        //Semaphore
        RedissonClient redisson = Redisson.create(config);
        final RSemaphore semaphore = redisson.getSemaphore("semaphore");
        semaphore.trySetPermits(3);
        for (int i = 0; i < 10; i++) {
            new Thread(new Runnable() {
                public void run() {
                    try {
                        System.out.println(new Date() + "：线程[" + Thread.currentThread().getName() + "]尝试获取Semaphore锁");
                        semaphore.acquire();
                        System.out.println(new Date() + "：线程[" + Thread.currentThread().getName() + "]成功获取到了Semaphore锁，开始工作");
                        Thread.sleep(3000);
                        semaphore.release();
                        System.out.println(new Date() + "：线程[" + Thread.currentThread().getName() + "]释放Semaphore锁");
                    } catch (Exception e) {
                        e.printStackTrace();
                    }
                }
            }).start();
        }
    }
}

3.Redisson的Semaphore源码剖析

(1)Semaphore的初始化

(2)Semaphore设置允许获取的锁数量

(3)客户端尝试获取Semaphore的锁

(4)客户端释放Semaphore的锁

(1)Semaphore的初始化

public class Redisson implements RedissonClient {
    //Redis的连接管理器，封装了一个Config实例
    protected final ConnectionManager connectionManager;
    //Redis的命令执行器，封装了一个ConnectionManager实例
    protected final CommandAsyncExecutor commandExecutor;
    ...
    protected Redisson(Config config) {
        this.config = config;
        Config configCopy = new Config(config);
        //初始化Redis的连接管理器
        connectionManager = ConfigSupport.createConnectionManager(configCopy);
        ...  
        //初始化Redis的命令执行器
        commandExecutor = new CommandSyncService(connectionManager, objectBuilder);
        ...
    }

    @Override
    public RSemaphore getSemaphore(String name) {
        return new RedissonSemaphore(commandExecutor, name);
    }
    ...
}

public class RedissonSemaphore extends RedissonExpirable implements RSemaphore {
    private final SemaphorePubSub semaphorePubSub;
    final CommandAsyncExecutor commandExecutor;

    public RedissonSemaphore(CommandAsyncExecutor commandExecutor, String name) {
        super(commandExecutor, name);
        this.commandExecutor = commandExecutor;
        this.semaphorePubSub = commandExecutor.getConnectionManager().getSubscribeService().getSemaphorePubSub();
    }
    ...
}

(2)Semaphore设置允许获取的锁数量

public class RedissonSemaphore extends RedissonExpirable implements RSemaphore {
    ...
    @Override
    public boolean trySetPermits(int permits) {
        return get(trySetPermitsAsync(permits));
    }

    @Override
    public RFuture<Boolean> trySetPermitsAsync(int permits) {
        RFuture<Boolean> future = commandExecutor.evalWriteAsync(getRawName(), LongCodec.INSTANCE, RedisCommands.EVAL_BOOLEAN,
            //执行命令"get semaphore"，获取到当前的数值
            "local value = redis.call('get', KEYS[1]); " +
            "if (value == false) then " +
                //然后执行命令"set semaphore 3"
                //设置这个信号量允许客户端同时获取锁的总数量为3
                "redis.call('set', KEYS[1], ARGV[1]); " +
                "redis.call('publish', KEYS[2], ARGV[1]); " +
                "return 1;" +
            "end;" +
            "return 0;",
            Arrays.asList(getRawName(), getChannelName()),
            permits
        );

        if (log.isDebugEnabled()) {
            future.onComplete((r, e) -> {
                if (r) {
                    log.debug("permits set, permits: {}, name: {}", permits, getName());
                } else {
                    log.debug("unable to set permits, permits: {}, name: {}", permits, getName());
                }
            });
        }
        return future;
    }
    ...
}

首先执行命令"get semaphore"，获取到当前的数值。然后执行命令"set semaphore 3"，也就是设置这个信号量允许客户端同时获取锁的总数量为3。

(3)客户端尝试获取Semaphore的锁

public class RedissonSemaphore extends RedissonExpirable implements RSemaphore {
    ...
    private final SemaphorePubSub semaphorePubSub;
    final CommandAsyncExecutor commandExecutor;

    public RedissonSemaphore(CommandAsyncExecutor commandExecutor, String name) {
        super(commandExecutor, name);
        this.commandExecutor = commandExecutor;
        this.semaphorePubSub = commandExecutor.getConnectionManager().getSubscribeService().getSemaphorePubSub();
    }

    @Override
    public void acquire() throws InterruptedException {
        acquire(1);
    }
    
    @Override
    public void acquire(int permits) throws InterruptedException {
        if (tryAcquire(permits)) {
            return;
        }
        CompletableFuture<RedissonLockEntry> future = subscribe();
        commandExecutor.syncSubscriptionInterrupted(future);
        try {
            while (true) {
                if (tryAcquire(permits)) {
                    return;
                }
                //获取Redisson的Semaphore失败，于是便调用本地JDK的Semaphore的acquire()方法，此时当前线程会被阻塞
                //之后如果Redisson的Semaphore释放了锁，那么当前客户端便会通过监听订阅事件释放本地JDK的Semaphore，唤醒被阻塞的线程，继续执行while循环
                //注意：getLatch()返回的是JDK的Semaphore = "new Semaphore(0)" ==> (state - permits)


                //首先调用CommandAsyncService.getNow()方法
                //然后调用RedissonLockEntry.getLatch()方法
                //接着调用JDK的Semaphore的acquire()方法
                commandExecutor.getNow(future).getLatch().acquire();
            }
        } finally {
            unsubscribe(commandExecutor.getNow(future));
        }
    }
    
    @Override
    public boolean tryAcquire(int permits) {
        //异步转同步
        return get(tryAcquireAsync(permits));
    }
    
    @Override
    public RFuture<Boolean> tryAcquireAsync(int permits) {
        if (permits < 0) {
            throw new IllegalArgumentException("Permits amount can't be negative");
        }
        if (permits == 0) {
            return RedissonPromise.newSucceededFuture(true);
        }
        return commandExecutor.evalWriteAsync(getRawName(), LongCodec.INSTANCE, RedisCommands.EVAL_BOOLEAN,
            //执行命令"get semaphore"，获取到当前值
            "local value = redis.call('get', KEYS[1]); "+
            //如果semaphore的当前值不是false，且大于客户端线程申请获取锁的数量
            "if (value ~= false and tonumber(value) >= tonumber(ARGV[1])) then " +
                //执行"decrby semaphore 1"，将信号量允许获取锁的总数量递减1
                "local val = redis.call('decrby', KEYS[1], ARGV[1]); " +
                "return 1; " +
            "end; " +
            //如果semaphore的值变为0，那么客户端就无法获取锁了，此时返回false
            "return 0;",
            Collections.<Object>singletonList(getRawName()),
            permits//ARGV[1]默认是1
        );
    }
    ...
}

public class CommandAsyncService implements CommandAsyncExecutor {
    ...
    @Override
    public <V> V getNow(CompletableFuture<V> future) {
        try {
            return future.getNow(null);
        } catch (Exception e) {
            return null;
        }
    }
    ...
}

public class RedissonLockEntry implements PubSubEntry<RedissonLockEntry> {
    private final Semaphore latch;
    ...
    public RedissonLockEntry(CompletableFuture<RedissonLockEntry> promise) {
        super();
        this.latch = new Semaphore(0);
        this.promise = promise;
    }
    
    public Semaphore getLatch() {
        return latch;
    }
    ...
}

执行命令"get semaphore"，获取到semaphore的当前值。如果semaphore的当前值不是false，且大于客户端线程申请获取锁的数量。那么就执行"decrby semaphore 1"，将信号量允许获取锁的总数量递减1。

如果semaphore的值变为0，那么客户端就无法获取锁了，此时tryAcquire()方法返回false。表示获取semaphore的锁失败了，于是当前客户端线程便会通过本地JDK的Semaphore进行阻塞。

当客户端后续收到一个订阅事件把本地JDK的Semaphore进行释放后，便会唤醒阻塞线程继续while循环。在while循环中，会不断尝试获取这个semaphore的锁，如此循环往复，直到成功获取。

(4)客户端释放Semaphore的锁

public class RedissonSemaphore extends RedissonExpirable implements RSemaphore {
    ...
    @Override
    public void release() {
        release(1);
    }

    @Override
    public void release(int permits) {
        get(releaseAsync(permits));
    }

    @Override
    public RFuture<Void> releaseAsync(int permits) {
        if (permits < 0) {
            throw new IllegalArgumentException("Permits amount can't be negative");
        }
        if (permits == 0) {
            return RedissonPromise.newSucceededFuture(null);
        }

        RFuture<Void> future = commandExecutor.evalWriteAsync(getRawName(), StringCodec.INSTANCE, RedisCommands.EVAL_VOID,
            //执行命令"incrby semaphore 1"
            "local value = redis.call('incrby', KEYS[1], ARGV[1]); " +
            "redis.call('publish', KEYS[2], value); ",
            Arrays.asList(getRawName(), getChannelName()),
            permits
        );
        if (log.isDebugEnabled()) {
            future.onComplete((o, e) -> {
                if (e == null) {
                    log.debug("released, permits: {}, name: {}", permits, getName());
                }
            });
        }
        return future;
    }
    ...
}

//订阅semaphore不为0的事件，semaphore不为0时会触发执行这里的监听回调
public class SemaphorePubSub extends PublishSubscribe<RedissonLockEntry> {
    public SemaphorePubSub(PublishSubscribeService service) {
        super(service);
    }
    
    @Override
    protected RedissonLockEntry createEntry(CompletableFuture<RedissonLockEntry> newPromise) {
        return new RedissonLockEntry(newPromise);
    }
    
    @Override
    protected void onMessage(RedissonLockEntry value, Long message) {
    Runnable runnableToExecute = value.getListeners().poll();
        if (runnableToExecute != null) {
            runnableToExecute.run();
        }
        //将客户端本地JDK的Semaphore进行释放
        value.getLatch().release(Math.min(value.acquired(), message.intValue()));
    }
}

//订阅锁被释放的事件，锁被释放为0时会触发执行这里的监听回调
public class LockPubSub extends PublishSubscribe<RedissonLockEntry> {
    public static final Long UNLOCK_MESSAGE = 0L;
    public static final Long READ_UNLOCK_MESSAGE = 1L;
    
    public LockPubSub(PublishSubscribeService service) {
        super(service);
    }  
      
    @Override
    protected RedissonLockEntry createEntry(CompletableFuture<RedissonLockEntry> newPromise) {
        return new RedissonLockEntry(newPromise);
    }
    
    @Override
    protected void onMessage(RedissonLockEntry value, Long message) {
        if (message.equals(UNLOCK_MESSAGE)) {
            Runnable runnableToExecute = value.getListeners().poll();
            if (runnableToExecute != null) {
                runnableToExecute.run();
            }
            value.getLatch().release();
        } else if (message.equals(READ_UNLOCK_MESSAGE)) {
            while (true) {
                Runnable runnableToExecute = value.getListeners().poll();
                if (runnableToExecute == null) {
                    break;
                }
                runnableToExecute.run();
            }
            //将客户端本地JDK的Semaphore进行释放
            value.getLatch().release(value.getLatch().getQueueLength());
        }
    }
}

客户端释放Semaphore的锁时，会执行命令"incrby semaphore 1"。每当客户端释放掉permits个锁，就会将信号量的值累加permits，这样Semaphore信号量的值就不再是0了。然后通过publish命令发布一个事件，之后订阅了该事件的其他客户端都会对getLatch()返回的本地JDK的Semaphore进行加1。于是其他客户端正在被本地JDK的Semaphore进行阻塞的线程，就会被唤醒继续执行。此时其他客户端就可以尝试获取到这个信号量的锁，然后再次将这个Semaphore的值递减1。

4.Redisson的CountDownLatch简介

(1)Redisson的CountDownLatch原理图解

(2)Redisson的CountDownLatch使用演示

(1)Redisson的CountDownLatch原理图解

CountDownLatch的基本原理：要求必须有n个线程来进行countDown，才能让执行await的线程继续执行。如果没有达到指定数量的线程来countDown，会导致执行await的线程阻塞。

(2)Redisson的CountDownLatch使用演示

public class RedissonDemo {
    public static void main(String[] args) throws Exception {
        //连接3主3从的Redis CLuster
        Config config = new Config();
        ...
        //CountDownLatch
        final RedissonClient redisson = Redisson.create(config);
        RCountDownLatch latch = redisson.getCountDownLatch("myCountDownLatch");
        //1.设置可以countDown的数量为3
        latch.trySetCount(3);
        System.out.println(new Date() + "：线程[" + Thread.currentThread().getName() + "]设置了必须有3个线程执行countDown，进入等待中。。。");

        for (int i = 0; i < 3; i++) {
            new Thread(new Runnable() {
                public void run() {
                    try {
                        System.out.println(new Date() + "：线程[" + Thread.currentThread().getName() + "]在做一些操作，请耐心等待。。。。。。");
                        Thread.sleep(3000);
                        RCountDownLatch localLatch = redisson.getCountDownLatch("myCountDownLatch");
                        localLatch.countDown();
                        System.out.println(new Date() + "：线程[" + Thread.currentThread().getName() + "]执行countDown操作");
                    } catch (Exception e) {
                        e.printStackTrace();
                    }
                }
            }).start();
        }
        latch.await();
        System.out.println(new Date() + "：线程[" + Thread.currentThread().getName() + "]收到通知，有3个线程都执行了countDown操作，可以继续往下执行");
    }
}

5.Redisson的CountDownLatch源码剖析

(1)CountDownLatch的初始化

(2)trySetCount()方法设置countDown的数量

(3)awati()方法进行阻塞等待

(4)countDown()方法对countDown的数量递减

(1)CountDownLatch的初始化

public class Redisson implements RedissonClient {
    //Redis的连接管理器，封装了一个Config实例
    protected final ConnectionManager connectionManager;
    //Redis的命令执行器，封装了一个ConnectionManager实例
    protected final CommandAsyncExecutor commandExecutor;
    ...
    protected Redisson(Config config) {
        this.config = config;
        Config configCopy = new Config(config);
        //初始化Redis的连接管理器
        connectionManager = ConfigSupport.createConnectionManager(configCopy);
        ...  
        //初始化Redis的命令执行器
        commandExecutor = new CommandSyncService(connectionManager, objectBuilder);
        ...
    }

    @Override
    public RCountDownLatch getCountDownLatch(String name) {
        return new RedissonCountDownLatch(commandExecutor, name);
    }
    ...
}

public class RedissonCountDownLatch extends RedissonObject implements RCountDownLatch {
    ...
    private final CountDownLatchPubSub pubSub;
    private final String id;
    protected RedissonCountDownLatch(CommandAsyncExecutor commandExecutor, String name) {
        super(commandExecutor, name);
        this.id = commandExecutor.getConnectionManager().getId();
        this.pubSub = commandExecutor.getConnectionManager().getSubscribeService().getCountDownLatchPubSub();
    }
    ...
}

(2)trySetCount()方法设置countDown的数量

trySetCount()方法的工作就是执行命令"set myCountDownLatch 3"。

public class RedissonCountDownLatch extends RedissonObject implements RCountDownLatch {
    ...
    @Override
    public boolean trySetCount(long count) {
        return get(trySetCountAsync(count));
    }

    @Override
    public RFuture<Boolean> trySetCountAsync(long count) {
        return commandExecutor.evalWriteAsync(getRawName(), LongCodec.INSTANCE, RedisCommands.EVAL_BOOLEAN,
            "if redis.call('exists', KEYS[1]) == 0 then " +
                "redis.call('set', KEYS[1], ARGV[2]); " +
                "redis.call('publish', KEYS[2], ARGV[1]); " +
                "return 1 " +
            "else " +
                "return 0 " +
            "end",
            Arrays.asList(getRawName(), getChannelName()),
            CountDownLatchPubSub.NEW_COUNT_MESSAGE,
            count
        );
    }
    ...
}

(3)awati()方法进行阻塞等待

public class RedissonCountDownLatch extends RedissonObject implements RCountDownLatch {
    ...
    @Override
    public void await() throws InterruptedException {
        if (getCount() == 0) {
            return;
        }
        CompletableFuture<RedissonCountDownLatchEntry> future = subscribe();
        try {
            commandExecutor.syncSubscriptionInterrupted(future);
            while (getCount() > 0) {
                // waiting for open state
                //获取countDown的数量还大于0，就先阻塞线程，然后再等待唤醒，执行while循环
                //其中getLatch()返回的是JDK的semaphore = "new Semaphore(0)" ==> (state - permits)
                commandExecutor.getNow(future).getLatch().await();
            }
        } finally {
            unsubscribe(commandExecutor.getNow(future));
        }
    }

    @Override
    public long getCount() {
        return get(getCountAsync());
    }

    @Override
    public RFuture<Long> getCountAsync() {
        //执行命令"get myCountDownLatch"
        return commandExecutor.writeAsync(getRawName(), LongCodec.INSTANCE, RedisCommands.GET_LONG, getRawName());
    }
    ...
}

在while循环中，首先会执行命令"get myCountDownLatch"去获取countDown值。如果该值不大于0，就退出循环不阻塞线程。如果该值大于0，则说明还没有指定数量的线程去执行countDown操作，于是就会先阻塞线程，然后再等待唤醒来继续循环。

(4)countDown()方法对countDown的数量递减

public class RedissonCountDownLatch extends RedissonObject implements RCountDownLatch {
    ...
    @Override
    public void countDown() {
        get(countDownAsync());
    }

    @Override
    public RFuture<Void> countDownAsync() {
        return commandExecutor.evalWriteNoRetryAsync(getRawName(), LongCodec.INSTANCE, RedisCommands.EVAL_BOOLEAN,
            "local v = redis.call('decr', KEYS[1]);" +
            "if v <= 0 then redis.call('del', KEYS[1]) end;" +
            "if v == 0 then redis.call('publish', KEYS[2], ARGV[1]) end;",
            Arrays.<Object>asList(getRawName(), getChannelName()),
            CountDownLatchPubSub.ZERO_COUNT_MESSAGE
        );
    }
    ...
}

countDownAsync()方法会执行decr命令，将countDown的数量进行递减1。如果这个值已经小于等于0，就执行del命令删除掉该CoutDownLatch。如果是这个值为0，还会发布一条消息：

publish redisson_countdownlatch__channel__{anyCountDownLatch} 0