分布式计数器的思路是:指定一个Zookeeper数据节点作为计数器,多个应用实例在分布式锁的控制下,通过更新该数据节点的内容来实现计数功能。
Curator中封装了实现,例如 DistributedAtomicInteger 和 DistributedAtomicLong。
以下是我写的一个测试用例:
- package com.my.CuratorTest;
- import org.apache.curator.framework.CuratorFramework;
- import org.apache.curator.framework.CuratorFrameworkFactory;
- import org.apache.curator.framework.recipes.atomic.AtomicValue;
- import org.apache.curator.framework.recipes.atomic.DistributedAtomicLong;
- import org.apache.curator.retry.ExponentialBackoffRetry;
- import org.apache.curator.retry.RetryNTimes;
- import java.util.concurrent.CyclicBarrier;
- import java.util.concurrent.ExecutorService;
- import java.util.concurrent.Executors;
- /**
- * Title: 分布式计数器演示<br/>
- * Intention: <br/>
- * <p>
- * Class Name: com.my.CuratorTest.RecipesDisAutomicLong<br/>
- * Create Date: 2017/8/20 22:48 <br/>
- * Project Name: MyTest <br/>
- * Company: All Rights Reserved. <br/>
- * Copyright © 2017 <br/>
- * </p>
- * <p>
- * author: GaoWei <br/>
- * 1st_examiner: <br/>
- * 2nd_examiner: <br/>
- * </p>
- *
- * @version 1.0
- * @since JDK 1.7
- */
- public class RecipesDisAutomicLong {
- static String disAutomicIntPath = "/curator_recipes_distatomicint_path3";
- static CuratorFramework client = CuratorFrameworkFactory.builder()
- .connectString("127.0.0.1:2181")
- .retryPolicy(new ExponentialBackoffRetry(1000, 3)).build();
- static DistributedAtomicLong atomicLong =
- new DistributedAtomicLong(client, disAutomicIntPath, new RetryNTimes(10, 500),
- null);
- public static void main(String[] args) throws Exception{
- client.start();
- Long[] nums = {1L, 2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L};
- ExecutorService executor = Executors.newFixedThreadPool(nums.length);
- CyclicBarrier barrier = new CyclicBarrier(nums.length);
- atomicLong.compareAndSet(atomicLong.get().postValue(), new Long(0));
- for (int i=0;i<nums.length;i++) {
- final int k = i;
- executor.execute(new Runnable() {
- @Override
- public void run() {
- try {
- barrier.await();
- System.out.println(Thread.currentThread().getName()+" " + System.nanoTime()+ " , 开始执行");
- AtomicValue<Long> av = atomicLong.add(nums[k]);
- System.out.println("to add value=" + nums[k] + ", Result=" + av.succeeded() + " preValue=" + av.preValue()
- + " postValue=" + av.postValue());
- } catch (Exception e) {
- e.printStackTrace();
- }
- }
- });
- }
- executor.shutdown();
- }
- }
运行结果:
pool-3-thread-10 89586635189009 , 开始执行
pool-3-thread-8 89586635508120 , 开始执行
pool-3-thread-9 89586635501453 , 开始执行
pool-3-thread-7 89586635493898 , 开始执行
pool-3-thread-6 89586635480564 , 开始执行
pool-3-thread-5 89586635470342 , 开始执行
pool-3-thread-4 89586635427231 , 开始执行
pool-3-thread-3 89586635410787 , 开始执行
pool-3-thread-2 89586635360564 , 开始执行
pool-3-thread-1 89586635236564 , 开始执行
to add value=10, Result=true preValue=0 postValue=10
to add value=2, Result=true preValue=10 postValue=12
to add value=6, Result=true preValue=12 postValue=18
to add value=1, Result=true preValue=18 postValue=19
to add value=7, Result=true preValue=19 postValue=26
to add value=3, Result=true preValue=26 postValue=29
to add value=4, Result=true preValue=29 postValue=33
to add value=8, Result=true preValue=33 postValue=41
to add value=9, Result=true preValue=41 postValue=50
to add value=5, Result=true preValue=50 postValue=55
如果在DistributedAtomicLong的构造方法参数中,RetryNTimes重试次数不够,比如是3,你会发现并不一定每次加数都会成功。显然这里是用了乐观锁机制,它并不保证操作一定成功(它在重试这么多次中都没有成功获得锁,导致操作没有执行),所以我们有必要通过调用 av.succeeded() 来查看此次加数是否成功。
下面是RetryNTimes为3时的某一次运行结果:
pool-3-thread-1 89922027531135 , 开始执行
pool-3-thread-5 89922027681802 , 开始执行
pool-3-thread-8 89922027737357 , 开始执行
pool-3-thread-4 89922027673802 , 开始执行
pool-3-thread-9 89922028120024 , 开始执行
pool-3-thread-10 89922027531580 , 开始执行
pool-3-thread-2 89922027616024 , 开始执行
pool-3-thread-3 89922027606246 , 开始执行
pool-3-thread-7 89922027722691 , 开始执行
pool-3-thread-6 89922027699580 , 开始执行
to add value=9, Result=true preValue=0 postValue=9
to add value=10, Result=true preValue=9 postValue=19
to add value=4, Result=true preValue=19 postValue=23
to add value=7, Result=true preValue=23 postValue=30
to add value=3, Result=true preValue=30 postValue=33
to add value=2, Result=true preValue=33 postValue=35
to add value=5, Result=true preValue=35 postValue=40
to add value=1, Result=false preValue=35 postValue=0
to add value=6, Result=false preValue=35 postValue=0
to add value=8, Result=false preValue=35 postValue=0
参考:
1、从PAXOS到ZOOKEEPER分布式一致性原理与实践
本文参考链接:https://www.cnblogs.com/maohuidong/p/8407498.html