java并发实战:连接池实现

池化技术简介

在我们使用数据库的过程中,我们往往使用数据库连接池而不是直接使用数据库连接进行操作,这是因为每一个数据库连接的创建和销毁的代价是昂贵的,而池化技术则预先创建了资源,这些资源是可复用的,这样就保证了在多用户情况下只能使用指定数目的资源,避免了一个用户创建一个连接资源,造成程序运行开销过大。关于Java并发编程的总结和思考

连接池实现原理

这里只实现一个简易的连接池,更多复杂的需求可根据该连接池进行改进,该连接池主要参数如下:

  1. 一个繁忙队列busy
  2. 一个空闲队列idle
  3. 连接池最大活动连接数maxActive
  4. 连接池最大等待时间maxWait
  5. 连接池的活动连接数activeSize

程序流程图如下:

在这里插入图片描述

代码实现

泛型接口ConnectionPool.java


public interface ConnectionPool<T> {

    /**
     * 初始化池资源
     * @param maxActive 池中最大活动连接数
     * @param maxWait 最大等待时间
     */
    void init(Integer maxActive, Long maxWait);

    /**
     * 从池中获取资源
     * @return 连接资源
     */
    T getResource() throws Exception;

    /**
     * 释放连接
     * @param connection 正在使用的连接
     */
    void release(T connection) throws Exception;

    /**
     * 释放连接池资源
     */
    void close();


}

以zookeeper为例,实现zookeeper连接池,ZookeeperConnectionPool.java


public class ZookeeperConnectionPool implements ConnectionPool<ZooKeeper> {
    //最大活动连接数
    private Integer maxActive; 
    //最大等待时间
    private Long maxWait; 
    //空闲队列
    private LinkedBlockingQueue<ZooKeeper> idle = new LinkedBlockingQueue<>();
    //繁忙队列
    private LinkedBlockingQueue<ZooKeeper> busy = new LinkedBlockingQueue<>();
    //连接池活动连接数
    private AtomicInteger activeSize = new AtomicInteger(0);
    //连接池关闭标记
    private AtomicBoolean isClosed = new AtomicBoolean(false);
    //总共获取的连接记数
    private AtomicInteger createCount = new AtomicInteger(0);
    //等待zookeeper客户端创建完成的计数器
    private static ThreadLocal<CountDownLatch> latchThreadLocal = ThreadLocal.withInitial(() -> new CountDownLatch(1));

    public ZookeeperConnectionPool(Integer maxActive, Long maxWait) {
        this.init(maxActive, maxWait);
    }

    @Override
    public void init(Integer maxActive, Long maxWait) {
        this.maxActive = maxActive;
        this.maxWait = maxWait;
    }

    @Override
    public ZooKeeper getResource() throws Exception {
        ZooKeeper zooKeeper;
        Long nowTime = System.currentTimeMillis();
        final CountDownLatch countDownLatch = latchThreadLocal.get();
        
        //空闲队列idle是否有连接
        if ((zooKeeper = idle.poll()) == null) {
            //判断池中连接数是否小于maxActive
            if (activeSize.get() < maxActive) {
                //先增加池中连接数后判断是否小于等于maxActive
                if (activeSize.incrementAndGet() <= maxActive) {
                    //创建zookeeper连接
                    zooKeeper = new ZooKeeper("localhost", 5000, (watch) -> {
                        if (watch.getState() == Watcher.Event.KeeperState.SyncConnected) {
                            countDownLatch.countDown();
                        }
                    });
                    countDownLatch.await();
                    System.out.println("Thread:" + Thread.currentThread().getId() + "获取连接:" + createCount.incrementAndGet() + "条");
                    busy.offer(zooKeeper);
                    return zooKeeper;
                } else {
                    //如增加后发现大于maxActive则减去增加的
                    activeSize.decrementAndGet();
                }
            }
            //若活动线程已满则等待busy队列释放连接
            try {
                System.out.println("Thread:" + Thread.currentThread().getId() + "等待获取空闲资源");
                Long waitTime = maxWait - (System.currentTimeMillis() - nowTime);
                zooKeeper = idle.poll(waitTime, TimeUnit.MILLISECONDS);
            } catch (InterruptedException e) {
                throw new Exception("等待异常");
            }
            //判断是否超时
            if (zooKeeper != null) {
                System.out.println("Thread:" + Thread.currentThread().getId() + "获取连接:" + createCount.incrementAndGet() + "条");
                busy.offer(zooKeeper);
                return zooKeeper;
            } else {
                System.out.println("Thread:" + Thread.currentThread().getId() + "获取连接超时,请重试!");
                throw new Exception("Thread:" + Thread.currentThread().getId() + "获取连接超时,请重试!");
            }
        }
        //空闲队列有连接,直接返回
        busy.offer(zooKeeper);
        return zooKeeper;
    }

    @Override
    public void release(ZooKeeper connection) throws Exception {
        if (connection == null) {
            System.out.println("connection 为空");
            return;
        }
        if (busy.remove(connection)){
            idle.offer(connection);
        } else {
            activeSize.decrementAndGet();
            throw new Exception("释放失败");
        }
    }

    @Override
    public void close() {
        if (isClosed.compareAndSet(false, true)) {
            idle.forEach((zooKeeper) -> {
                try {
                    zooKeeper.close();
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
            });
            busy.forEach((zooKeeper) -> {
                try {
                    zooKeeper.close();
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
            });
        }
    }
}

测试用例

这里创建20个线程并发测试连接池,Test.java


public class Test {

    public static void main(String[] args) throws Exception {
        int threadCount = 20;
        Integer maxActive = 10;
        Long maxWait = 10000L;
        ZookeeperConnectionPool pool = new ZookeeperConnectionPool(maxActive, maxWait);
        CountDownLatch countDownLatch = new CountDownLatch(20);
        for (int i = 0; i < threadCount; i++) {
            new Thread(() -> {
                countDownLatch.countDown();
                try {
                    countDownLatch.await();
                    ZooKeeper zooKeeper = pool.getResource();
                    Thread.sleep(2000);
                    pool.release(zooKeeper);
                } catch (InterruptedException e) {
                    e.printStackTrace();
                } catch (Exception e) {
                    e.printStackTrace();
                }

            }).start();
        }
        while (true){

        }
    }
}

来源:https://segmentfault.com/a/1190000017926611

posted @ 2019-01-18 11:50  sfornt  阅读(5945)  评论(0编辑  收藏  举报