java 使用zookeeper包实现zookeeper分布式锁

zookeeper不可重入锁的实现代码

import org.apache.zookeeper.*;
import org.apache.zookeeper.data.Stat;
import java.io.IOException;
import java.util.Collections;
import java.util.List;

public class DistributedLockExample {
    private static final String ZOOKEEPER_CONNECTION_STRING = "localhost:2181";
    private static final int SESSION_TIMEOUT = 5000;
    private static final String LOCK_PATH = "/distributed_lock";
    private ZooKeeper zooKeeper;
    //当前最新的一个节点
    private String currentZnodeName;

    public DistributedLockExample() throws IOException {
        this.zooKeeper = new ZooKeeper(ZOOKEEPER_CONNECTION_STRING, SESSION_TIMEOUT, new Watcher() {
            @Override
            public void process(WatchedEvent event) {
                // Watcher实现
            }
        });
    }

    public void acquireLock() throws KeeperException, InterruptedException {
        while (true) {
            createLockNode();
            //获取锁节点下的所有子节点
            List<String> children = zooKeeper.getChildren(LOCK_PATH, false);
            //排序
            Collections.sort(children);
            //获取小的节点，也就是最先创建的那个节点
            String smallestNode = children.get(0);
            //如果当前最新节点等于最小节点，那么就代表获取到了锁
            if (currentZnodeName.equals(LOCK_PATH + "/" + smallestNode)) {
                return;
            } else {
                //这里等待获取锁，利用CountDownLatch， 是 Java 中的一个同步辅助类
                //CountDownLatch的countDown()可以把计数器减一， await()会让线程一直等待执行
                //直到计数器为0的时候
//children.get(Collections.binarySearch(children, currentZnodeName.substring(LOCK_PATH.length() + 1)) - 1)这段代码是获取当前最新节点的前一个节点，然后只监听前一个节点可以避免惊群效应。
                waitForLock(children.get(Collections.binarySearch(children, currentZnodeName.substring(LOCK_PATH.length() + 1)) - 1));
            }
        }
    }

    private void createLockNode() throws KeeperException, InterruptedException {
        Stat stat = zooKeeper.exists(LOCK_PATH, false);
        if (stat == null) {
            //如果stat锁的节点不存在，则创建一个持久的锁节点
            zooKeeper.create(LOCK_PATH, new byte[0], ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.PERSISTENT);
        }
        //在指定的锁节点下创建一个临时的顺序节点，并且赋值给currentZnodeName（当前最新的一个节点）
        currentZnodeName = zooKeeper.create(LOCK_PATH + "/", new byte[0], ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.EPHEMERAL_SEQUENTIAL);
    }

    private void waitForLock(String prevZnodeName) throws KeeperException, InterruptedException {
        CountDownLatch latch = new CountDownLatch(1);
        Stat stat = zooKeeper.exists(LOCK_PATH + "/" + prevZnodeName, new Watcher() {
            @Override
            public void process(WatchedEvent event) {
                //如果监听的节点被删除，那么就会触发这个if，
                if (event.getType() == Event.EventType.NodeDeleted) {
                    //减一，等于零，那么下面latch.await()就会让线程不在等待，继续执行，然后就会重新进入acquireLock()的while (true)循环，来获取锁
                    latch.countDown();
                }
            }
        });
        if (stat != null) {
            latch.await();
        }
    }

    public void releaseLock() throws KeeperException, InterruptedException {
        zooKeeper.delete(currentZnodeName, -1);
        zooKeeper.close();
    }

    public static void main(String[] args) throws IOException, KeeperException, InterruptedException {
        DistributedLockExample lockExample = new DistributedLockExample();
        lockExample.acquireLock();
        // 执行需要加锁的业务逻辑
        lockExample.releaseLock();
    }
}

可重入锁的实现原理就是维护一个线程id和一个计数器，当同一个线程获取到最小节点时，可以拿到锁，当同一个线程的A方法调用B方法，A方法和B方法需要同一把锁时，通过判断线程id获取到锁，并且给锁的计数器加1，当B方法执行完时，给计数器减一，计数器达到0时，释放锁（删除节点）

import org.apache.zookeeper.*;
import org.apache.zookeeper.data.Stat;
import java.util.Collections;
import java.util.List;

public class ZookeeperReentrantLock {
    private ZooKeeper zk;
    private final String lockPath = "/reentrantLock";
    //利用ThreadLocal特性提供线程局部变量的存取。每个线程可以访问自己内部的 ThreadLocal 变量副本，互不影响，使用ThreadLocal要小心内存泄漏问题，所以用完后要手动删除ThreadLocal
    private ThreadLocal<String> lockNode = new ThreadLocal<>();
    private final int sessionTimeout = 3000; // 会话超时时间

    public ZookeeperReentrantLock(String hosts) throws Exception {
        zk = new ZooKeeper(hosts, sessionTimeout, (event) -> {
            if (event.getType() == Watcher.Event.EventType.NodeDeleted) {
                // 如果当前线程的锁节点被删除，可能是会话超时，需要重新创建
                acquireLock();
            }
        });
        ensureLockPath();
    }

    public void acquireLock() throws Exception {
        // 如果当前线程已经持有锁，则增加重入次数并返回
        String current = lockNode.get();
        if (current != null) {
            ((LockData) new Stat()).incrementAndGet();
            return;
        }

        // 创建临时顺序节点
        String createdPath = zk.create(lockPath + "/", new byte[0], ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.EPHEMERAL_SEQUENTIAL);
        List<String> children = zk.getChildren(lockPath, false);
        Collections.sort(children);

        // 尝试获取锁
        for (String child : children) {
            if (createdPath.equals(lockPath + "/" + child)) {
                // 当前线程创建的节点序号最小，获取锁
                lockNode.set(createdPath);
                break;
            } else {
                // 等待前一个节点释放锁
                zk.exists(lockPath + "/" + child, true);
            }
        }
    }

    public void releaseLock() throws Exception {
        String current = lockNode.get();
        if (current != null) {
            // 减少重入次数，如果为0，则释放锁
            LockData lockData = (LockData) zk.exists(current, false);
            if (lockData == null || lockData.decrementAndGet() == 0) {
                zk.delete(current, -1);
                lockNode.remove();
            }
        }
    }

    private void ensureLockPath() throws KeeperException, InterruptedException {
        Stat stat = zk.exists(lockPath, false);
        if (stat == null) {
            zk.create(lockPath, new byte[0], ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.PERSISTENT);
        }
    }

    private static class LockData extends Stat {
        private int count = 1;

        public void incrementAndGet() {
            count++;
        }

        public int decrementAndGet() {
            return --count;
        }
    }

    public static void main(String[] args) {
        try {
            ZookeeperReentrantLock lock = new ZookeeperReentrantLock("localhost:2181");
            lock.acquireLock();
            // 执行业务逻辑
            lock.releaseLock();
        } catch (Exception e) {
            e.printStackTrace();
        }
    }
}