从构建分布式秒杀系统聊聊分布式锁

从构建分布式秒杀系统聊聊分布式锁

1.案例介绍

在尝试了解分布式锁之前，大家可以想象一下，什么场景下会使用分布式锁？

单机应用架构中，秒杀案例使用ReentrantLcok或者synchronized来达到秒杀商品互斥的目的。然而在分布式系统中，会存在多台机器并行去实现同一个功能。也就是说，在多进程中，如果还使用以上JDK提供的进程锁，来并发访问数据库资源就可能会出现商品超卖的情况。因此，需要我们来实现自己的分布式锁。

实现一个分布式锁应该具备的特性：

• 高可用、高性能的获取锁与释放锁
• 在分布式系统环境下，一个方法或者变量同一时间只能被一个线程操作
• 具备锁失效机制，网络中断或宕机无法释放锁时，锁必须被删除，防止死锁
• 具备阻塞锁特性，即没有获取到锁，则继续等待获取锁
• 具备非阻塞锁特性，即没有获取到锁，则直接返回获取锁失败
• 具备可重入特性，一个线程中可以多次获取同一把锁，比如一个线程在执行一个带锁的方法，该方法中又调用了另一个需要相同锁的方法，则该线程可以直接执行调用的方法，而无需重新获得锁

在之前的秒杀案例中，我们曾介绍过关于分布式锁几种实现方式：

• 基于数据库实现分布式锁
• 基于 Redis 实现分布式锁
• 基于 Zookeeper 实现分布式锁

前两种对于分布式生产环境来说并不是特别推荐，高并发下数据库锁性能太差，Redis在锁时间限制和缓存一致性存在一定问题。这里我们重点介绍一下 Zookeeper 如何实现分布式锁。

2.实现原理

ZooKeeper是一个分布式的，开放源码的分布式应用程序协调服务，它内部是一个分层的文件系统目录树结构，规定同一个目录下只能存在唯一文件名。

ZooKeeper数据模型与文件系统目录树(源自网络)

3.数据模型

• PERSISTENT 持久化节点，节点创建后，不会因为会话失效而消失
• EPHEMERAL 临时节点， 客户端session超时此类节点就会被自动删除
• EPHEMERAL_SEQUENTIAL 临时自动编号节点
• PERSISTENT_SEQUENTIAL 顺序自动编号持久化节点，这种节点会根据当前已存在的节点数自动加 1

4.监视器（watcher）

当创建一个节点时，可以注册一个该节点的监视器，当节点状态发生改变时，watch被触发时，ZooKeeper将会向客户端发送且仅发送一条通知，因为watch只能被触发一次。

根据zookeeper的这些特性，我们来看看如何利用这些特性来实现分布式锁：

• 创建一个锁目录lock
• 线程A获取锁会在lock目录下，创建临时顺序节点
• 获取锁目录下所有的子节点，然后获取比自己小的兄弟节点，如果不存在，则说明当前线程顺序号最小，获得锁
• 线程B创建临时节点并获取所有兄弟节点，判断自己不是最小节点，设置监听(watcher)比自己次小的节点
• 线程A处理完，删除自己的节点，线程B监听到变更事件，判断自己是最小的节点，获得锁

5.代码分析

尽管ZooKeeper已经封装好复杂易出错的关键服务，将简单易用的接口和性能高效、功能稳定的系统提供给用户。但是如果让一个普通开发者去手撸一个分布式锁还是比较困难的，在秒杀案例中我们直接使用 Apache 开源的curator 开实现 Zookeeper 分布式锁。

这里我们使用以下版本，截止目前最新版4.0.1，maven依赖如下：

<dependency>
    <groupId>org.apache.curatorgroupId>
    <artifactId>curator-recipesartifactId>
    <version>2.12.0version>
dependency>

首先，我们看下InterProcessLock接口中的几个方法：

package org.apache.curator.framework.recipes.locks;

import java.util.concurrent.TimeUnit;

public interface InterProcessLock
{
    /**
     * Acquire the mutex - blocking until it's available. Each call to acquire must be balanced by a call
     * to {@link #release()}
     *
     * @throws Exception ZK errors, connection interruptions
     */
    public void acquire() throws Exception;

    /**
     * Acquire the mutex - blocks until it's available or the given time expires. Each call to acquire that returns true must be balanced by a call
     * to {@link #release()}
     *
     * @param time time to wait
     * @param unit time unit
     * @return true if the mutex was acquired, false if not
     * @throws Exception ZK errors, connection interruptions
     */
    public boolean acquire(long time, TimeUnit unit) throws Exception;

    /**
     * Perform one release of the mutex.
     *
     * @throws Exception ZK errors, interruptions, current thread does not own the lock
     */
    public void release() throws Exception;

    /**
     * Returns true if the mutex is acquired by a thread in this JVM
     *
     * @return true/false
     */
    boolean isAcquiredInThisProcess();
}

InterProcessLock接口总共有4个实现类，结构如下

• InterProcessMultiLock：管理多个锁的容器最为单一对象，当调用acquire方法获取锁时，所有的锁被同时获取，业务处理完成后，线程释放锁时也是同时释放所有的锁

  /**
   * A container that manages multiple locks as a single entity. When {@link #acquire()} is called,
   * all the locks are acquired. If that fails, any paths that were acquired are released. Similarly, when
   * {@link #release()} is called, all locks are released (failures are ignored).
   */
  public class InterProcessMultiLock implements InterProcessLock
  {
      private final List<InterProcessLock> locks;
  
      /**
       * Creates a multi lock of {@link InterProcessMutex}s
       *
       * @param client the client
       * @param paths list of paths to manage in the order that they are to be locked
       */
      public InterProcessMultiLock(CuratorFramework client, List<String> paths)
      {
          // paths get checked in each individual InterProcessMutex, so trust them here
          this(makeLocks(client, paths));
      }
  

• InterProcessMutex：支持可重入锁

  /**
   * A re-entrant mutex that works across JVMs. Uses Zookeeper to hold the lock. All processes in all JVMs that
   * use the same lock path will achieve an inter-process critical section. Further, this mutex is
   * "fair" - each user will get the mutex in the order requested (from ZK's point of view)
   */
  public class InterProcessMutex implements InterProcessLock, Revocable<InterProcessMutex>
  {
      private final LockInternals internals;
      private final String basePath;
  
      private final ConcurrentMap<Thread, LockData> threadData = Maps.newConcurrentMap();
  
      private static class LockData
      {
          final Thread owningThread;
          final String lockPath;
          final AtomicInteger lockCount = new AtomicInteger(1);
  
          private LockData(Thread owningThread, String lockPath)
          {
              this.owningThread = owningThread;
              this.lockPath = lockPath;
          }
      }
  
      private static final String LOCK_NAME = "lock-";
  

• InterProcessSemaphoreMutex : 不支持可重入锁

  /**
   * A NON re-entrant mutex that works across JVMs. Uses Zookeeper to hold the lock. All processes in all JVMs that
   * use the same lock path will achieve an inter-process critical section.
   */
  public class InterProcessSemaphoreMutex implements InterProcessLock
  {
      private final InterProcessSemaphoreV2 semaphore;
      private volatile Lease lease;
  
      /**
       * @param client the client
       * @param path path for the lock
       */
      public InterProcessSemaphoreMutex(CuratorFramework client, String path)
      {
          this.semaphore = new InterProcessSemaphoreV2(client, path, 1);
      }
  

• InternalInterProcessMutex：zk读写锁的静态内部类，充当zk读写锁

  public class InterProcessReadWriteLock
  {
      private final InterProcessMutex readMutex;
      private final InterProcessMutex writeMutex;
      
      private static class InternalInterProcessMutex extends InterProcessMutex
      {
          private final String lockName;
          private final byte[] lockData;
  
          InternalInterProcessMutex(CuratorFramework client, String path, String lockName, byte[] lockData, int maxLeases, LockInternalsDriver driver)
          {
              super(client, path, lockName, maxLeases, driver);
              this.lockName = lockName;
              this.lockData = lockData;
          }
  
          @Override
          public Collection<String> getParticipantNodes() throws Exception
          {
              Collection<String>  nodes = super.getParticipantNodes();
              Iterable<String>    filtered = Iterables.filter
              (
                  nodes,
                  new Predicate<String>()
                  {
                      @Override
                      public boolean apply(String node)
                      {
                          return node.contains(lockName);
                      }
                  }
              );
              return ImmutableList.copyOf(filtered);
          }
  
          @Override
          protected byte[] getLockNodeBytes()
          {
              return lockData;
          }
      }
  

这里以支持可重复锁InterProcessMutex针对源码展开介绍，了解zk充当分布式锁的锁机制，加锁和解锁过程

1)获取锁:

	/**
     * Acquire the mutex - blocking until it's available. Note: the same thread
     * can call acquire re-entrantly. Each call to acquire must be balanced by a call
     * to {@link #release()}
     *
     * @throws Exception ZK errors, connection interruptions
     */
    @Override
    public void acquire() throws Exception
    {
        if ( !internalLock(-1, null) )
        {
            throw new IOException("Lost connection while trying to acquire lock: " + basePath);
        }
    }

	/**
     * Acquire the mutex - blocks until it's available or the given time expires. Note: the same thread
     * can call acquire re-entrantly. Each call to acquire that returns true must be balanced by a call
     * to {@link #release()}
     *
     * @param time time to wait
     * @param unit time unit
     * @return true if the mutex was acquired, false if not
     * @throws Exception ZK errors, connection interruptions
     */
    @Override
    public boolean acquire(long time, TimeUnit unit) throws Exception
    {
        return internalLock(time, unit);
    }

具体实现：

	private boolean internalLock(long time, TimeUnit unit) throws Exception
    {
        /*
           Note on concurrency: a given lockData instance
           can be only acted on by a single thread so locking isn't necessary
        */

        Thread currentThread = Thread.currentThread();

        LockData lockData = threadData.get(currentThread);
        if ( lockData != null )
        {
            // re-entering
            lockData.lockCount.incrementAndGet();
            return true;
        }
		// 尝试获取zk锁
        String lockPath = internals.attemptLock(time, unit, getLockNodeBytes());
        if ( lockPath != null )
        {
            LockData newLockData = new LockData(currentThread, lockPath);
            threadData.put(currentThread, newLockData);
            return true;
        }
		//获取锁超时或者zk通信异常返回失败
        return false;
    }

Zookeeper获取锁实现：

	String attemptLock(long time, TimeUnit unit, byte[] lockNodeBytes) throws Exception
    {
        final long      startMillis = System.currentTimeMillis();
        final Long      millisToWait = (unit != null) ? unit.toMillis(time) : null;
        // 子节点标识
        final byte[]    localLockNodeBytes = (revocable.get() != null) ? new byte[0] : lockNodeBytes;
        // 重试次数
        int             retryCount = 0;

        String          ourPath = null;
        boolean         hasTheLock = false;
        boolean         isDone = false;
        // 自旋锁，循环获取锁
        while ( !isDone )
        {
            isDone = true;

            try
            {
                //在锁节点下创建临时且有序的子节点，例如:_c_008c1b07-d577-4e5f-8699-8f0f98a013b4-lock-000000001
                ourPath = driver.createsTheLock(client, path, localLockNodeBytes);
            //如果当前子节点序号最小，获得锁则直接返回，否则阻塞等待前一个子节点删除通知(release释放锁)
                hasTheLock = internalLockLoop(startMillis, millisToWait, ourPath);
            }
            catch ( KeeperException.NoNodeException e )
            {
                // gets thrown by StandardLockInternalsDriver when it can't find the lock node
                // this can happen when the session expires, etc. So, if the retry allows, just try it all again
                if ( client.getZookeeperClient().getRetryPolicy().allowRetry(retryCount++, System.currentTimeMillis() - startMillis, RetryLoop.getDefaultRetrySleeper()) )
                {
                    isDone = false;
                }
                else
                {
                    throw e;
                }
            }
        }

        if ( hasTheLock )
        {
            return ourPath;
        }

        return null;
    }

判断是否为最小节点：

	private boolean internalLockLoop(long startMillis, Long millisToWait, String ourPath) throws Exception
    {
        boolean     haveTheLock = false;
        boolean     doDelete = false;
        try
        {
            if ( revocable.get() != null )
            {
                client.getData().usingWatcher(revocableWatcher).forPath(ourPath);
            }

            // 自旋获取锁
            while ( (client.getState() == CuratorFrameworkState.STARTED) && !haveTheLock )
            {
                // 获取所有子节点集合
                List<String>        children = getSortedChildren();
                // 判断当前子节点是否是最小子节点
                String              sequenceNodeName = ourPath.substring(basePath.length() + 1); // +1 to include the slash

                PredicateResults    predicateResults = driver.getsTheLock(client, children, sequenceNodeName, maxLeases);
                if ( predicateResults.getsTheLock() )
                { // 当前节点是最小子节点，获得锁
                    haveTheLock = true;
                }
                else
                { // 获取前一个节点，用于监听
                    String  previousSequencePath = basePath + "/" + predicateResults.getPathToWatch();

                    synchronized(this)
                    {
                        try 
                        {
                            // use getData() instead of exists() to avoid leaving unneeded watchers which is a type of resource leak
                            client.getData().usingWatcher(watcher).forPath(previousSequencePath);
                            if ( millisToWait != null )
                            {
                                millisToWait -= (System.currentTimeMillis() - startMillis);
                                startMillis = System.currentTimeMillis();
                                if ( millisToWait <= 0 )
                                {
                                    doDelete = true;    // timed out - delete our node
                                    break;
                                }

                                wait(millisToWait);
                            }
                            else
                            {
                                wait();
                            }
                        }
                        catch ( KeeperException.NoNodeException e ) 
                        {
                            // it has been deleted (i.e. lock released). Try to acquire again
                           // 如果前一个子节点已经被删除则throw exception，只需要自旋获取一次即可
                        }
                    }
                }
            }
        }
        catch ( Exception e )
        {
            ThreadUtils.checkInterrupted(e);
            doDelete = true;
            throw e;
        }
        finally
        {
            if ( doDelete )
            {
                deleteOurPath(ourPath);
            }
        }
        return haveTheLock;
    }

2) 释放锁:

	/**
     * Perform one release of the mutex if the calling thread is the same thread that acquired it. If the
     * thread had made multiple calls to acquire, the mutex will still be held when this method returns.
     *
     * @throws Exception ZK errors, interruptions, current thread does not own the lock
     */
    @Override
    public void release() throws Exception
    {
        /*
            Note on concurrency: a given lockData instance
            can be only acted on by a single thread so locking isn't necessary
         */

        Thread currentThread = Thread.currentThread();
        LockData lockData = threadData.get(currentThread);
        if ( lockData == null )
        {
            throw new IllegalMonitorStateException("You do not own the lock: " + basePath);
        }

        // 获取重复梳理
        int newLockCount = lockData.lockCount.decrementAndGet();
        if ( newLockCount > 0 )
        {
            return;
        }
        if ( newLockCount < 0 )
        {
            throw new IllegalMonitorStateException("Lock count has gone negative for lock: " + basePath);
        }
        try
        {
            internals.releaseLock(lockData.lockPath);
        }
        finally
        {
            // 移除当前线程获取zk锁数据
            threadData.remove(currentThread);
        }
    }

6.测试案例

为了更好的理解其原理和代码分析中获取锁的过程，这里我们实现一个简单的Demo：

package com.netease.it.access.web.util;

import org.apache.curator.RetryPolicy;
import org.apache.curator.framework.CuratorFramework;
import org.apache.curator.framework.CuratorFrameworkFactory;
import org.apache.curator.framework.recipes.locks.InterProcessLock;
import org.apache.curator.framework.recipes.locks.InterProcessMutex;
import org.apache.curator.retry.ExponentialBackoffRetry;

import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;

/**
 * 基于curator的zookeeper分布式锁
 * Created by dujiayong on 2020/3/8.
 */
public class CuratorUtil {

    private static final String ZK_LOCK_DIR = "/curator/lock";

    public static void main(String[] args) {
        // 1:定义重试策略:初始睡眠1s 最大重试3次
        RetryPolicy retryPolicy = new ExponentialBackoffRetry(1000, 3);
        // 2.通过工厂创建连接
        CuratorFramework client = CuratorFrameworkFactory.newClient("localhost:2181", retryPolicy);
        // 3.打开连接
        client.start();

        // 4.分布式锁
        final InterProcessLock lock = new InterProcessMutex(client, ZK_LOCK_DIR);

        ExecutorService executorService = Executors.newFixedThreadPool(5);
        for (int i = 0; i < 5; i++) {
            executorService.submit(new Runnable() {
                @Override
                public void run() {
                    boolean acquire = false;
                    try {
                        acquire = lock.acquire(5000, TimeUnit.MILLISECONDS);
                        if (acquire) {
                            System.out.println("线程->" + Thread.currentThread().getName() + "获得锁成功");
                        } else {
                            System.out.println("线程->" + Thread.currentThread().getName() + "获得锁失败");
                        }
                        // 模拟业务，延迟4s
                        Thread.sleep(4000);
                    } catch (Exception e) {
                        e.printStackTrace();
                    } finally {
                        if (acquire) {
                            try {
                                lock.release();
                            } catch (Exception e) {
                                e.printStackTrace();
                            }
                        }
                    }
                }
            });
        }
    }

}

这里我们开启5个线程，每个线程获取锁的最大等待时间为5秒，为了模拟具体业务场景，方法中设置4秒等待时间。开始执行main方法，通过ZooInspector监控/curator/lock下的节点如下图：

对，没错，设置4秒的业务处理时长就是为了观察生成了几个顺序节点。果然如案例中所述，每个线程都会生成一个节点并且还是有序的。

观察控制台，我们会发现只有两个线程获取锁成功，另外三个线程超时获取锁失败会自动删除节点。线程执行完毕我们刷新一下/curator/lock节点，发现刚才创建的五个子节点已经不存在了。

---

来源：https://www.cnblogs.com/williamjie/p/9406031.html

zookeeper原理博客：

• zookeeper实现分布式锁（非公平锁）原理