Java分布式锁(一行代码搞定)

Java分布式锁(一行代码搞定)

前面几篇文章都是介绍的java juc lock包单机锁,但是目前很多应用都是分布式的(不同jvm进程),单机锁已经不能满足应用的需求了。
网上有关分布式锁的文章很多有基于memcached,redis,zookeeper,但感觉直接拿到线上使用,供公司内部使用还是差点火候,于是就一行代码搞定分布式锁文章就诞生了。

现在几乎每个应用都会用redis,我们就以redis为例使用分布式锁:
先看在springboot项目中使用方式:

lockTemplate.doBiz(LockedCallback<R> lockedCallback,String scene,String uniqueId,String key,Long expireSecond)

这样一来你只需要在回调类lockedCallback写你的业务逻辑,极大地简化了使用分布式锁的使用,能让你更加专注你的业务逻辑,集中管理所有的锁,而不是项目中每个人一套(而且并不是所有人都能用对)。

下面是我做的测试:

@RestController
@RequestMapping(path="/monitor")
public class MonitorController {

    @Autowired
    private LockTemplate<String> lockTemplate;
    private ExecutorService pool = Executors.newFixedThreadPool(10);
    /**
     * @description
     * @param
     * @return
     * @date: 2018/6/12 下午5:18
     * @author:yzMa
     */
    @GetMapping("/lock/go")
    public ApiResult<Object> lockTest(){
        for (int i = 0; i < 6; i++) {
            pool.submit(new Callable<String>() {
                @Override
                public String call() throws Exception {
                    return doSomething();
                }
            });
        }
        return ApiResult.buildSuccess(null);
    }

    /**
     * @description 竞态条件
     * @param 
     * @return 
     * @date: 2018/6/12 下午6:15
     * @author:yzMa
     */
    private String doSomething(){
        return lockTemplate.doBiz(new LockedCallback<String>() {
            @Override
            public String callback() {

                try {
        System.out.println(Thread.currentThread().getName()+" 成功获取到了锁开始执行逻辑");
                    Thread.sleep(4000);
                    System.out.println(Thread.currentThread().getName()+" 成功获取到了锁开始执行逻辑结束");
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
                return "aa";
            }
        },"testLock-scene","aaaa-uniqueId","aaaa",5L);
    }
}

受到网上前辈的影响,使用redis.set(key,val,expire,nx)命令加锁,使用lua脚本执行解锁,至于为什么自行百度下(多笔操作这个失败了,那个超时了,加锁了释放锁失败了如何处理等等),我也是踩过很多坑,分析过很多次才有了这个线上版本。

分布式锁模板类

@Slf4j
@Component
public class LockTemplate<R> {

    @Resource(name = "luaRedisDistributeLock")
    private DistributeLock distributeLock;
    public <R> R doBiz(LockedCallback<R> lockedCallback,String scene,String uniqueId,String key,Long expireSecond){
        if(StringUtils.isBlank(uniqueId)){
            log.info("|doBiz no uniqueId use serialize lock (变为分布式串行锁)");
            uniqueId = key+System.currentTimeMillis();
        }
        boolean acquiredSuccess = false;
        try{
            acquiredSuccess = distributeLock.tryLock(key, uniqueId, expireSecond);
            if(!acquiredSuccess){
                log.info("|doBiz acquire lock fail scene={}, key={},uniqueId={},expireSecond={}",scene,key,uniqueId,expireSecond);
                throw new BusinessException(ApiResult.ACQUIRE_LOCK_FAIL,ApiResult.ACQUIRE_LOCK_FAIL_MSG);
            }
            log.info("|doBiz acquire lock success scene={}, key={},uniqueId={},expireSecond={}",scene,key,uniqueId,expireSecond);
            return (R)lockedCallback.callback();

        }catch (Exception e){
            if(! (e instanceof BusinessException)){
                log.error("|doBiz|lockedCallback not BusinessException, scene={},key={},uniqueId={},expireSecond={} exception e:",scene,key,uniqueId,expireSecond,e);
            }
            throw e;
        }finally {
            if(acquiredSuccess){
                distributeLock.unlock(key,uniqueId);
                log.info("|doBiz release lock success scene={}, key={},uniqueId={},expireSecond={}",scene,key,uniqueId,expireSecond);
            }
        }
    }

    // 分布式锁模板,大锁 没有并发
    public <R> R doBiz(LockedCallback<R> lockedCallback,String scene,String key,Long expireSecond){

            return doBiz(lockedCallback,scene,"",key,expireSecond);
    }
}

lua分布式锁实现类

@Component("luaRedisDistributeLock")
public class LuaRedisDistributeLock implements DistributeLock{

    private String unlockLua = "if redis.call('get', KEYS[1]) == ARGV[1] then return redis.call('del', KEYS[1]) else return 0 end";

    @Autowired
    private StringRedisTemplate stringRedisTemplate;

    private JdkSerializationRedisSerializer jdkSerializer = new JdkSerializationRedisSerializer();

    class ExpirationSub extends Expiration{
        long expirationTime;
        TimeUnit timeUnit;
        public ExpirationSub(long expirationTime,TimeUnit timeUnit){
            super(expirationTime,timeUnit);
        }
    }

    @Override
    public void lock(String key, String uniqueId, Long expireTime) {
        throw new RuntimeException("lua redis not support for now");
    }

    @Override
    public boolean tryLock(String key,String uniqueId, Long expireSeconds) {
        Assert.notNull(key,"redis key 不能为空");
        Assert.notNull(uniqueId,"uniqueId 不能为空");
        Assert.notNull(expireSeconds,"expireTime 不能为空");
        Assert.isTrue(expireSeconds > 0 && expireSeconds <= 10 ,"锁的过期时间范围介于(0,10]秒");
        Boolean lockedSuccess = stringRedisTemplate.execute(new RedisCallback<Boolean>() {
            @Override
            public Boolean doInRedis(RedisConnection redisConnection) throws DataAccessException {
            return redisConnection.set(key.getBytes(), val, expirationSub, RedisStringCommands.SetOption.SET_IF_ABSENT);//老版本
            }
        });

        if(lockedSuccess){
            return true;
        }
        return false;
    }

    public void unlock(String key, String uniqueId) {

        //使用Lua脚本删除Redis中匹配value的key,可以避免由于方法执行时间过长而redis锁自动过期失效的时候误删其他线程的锁
        //spring自带的执行脚本方法中,集群模式直接抛出不支持执行脚本的异常,所以只能拿到原redis的connection来执行脚本
        stringRedisTemplate.execute(new RedisCallback<Boolean>() {
            @Override
            public Boolean doInRedis(RedisConnection redisConnection) throws DataAccessException {
                Object nativeConnection = redisConnection.getNativeConnection();

                //集群模式和单机模式虽然执行脚本的方法一样,但是没有共同的接口,所以只能分开执行
                Long luaResult = 0L;
                if (nativeConnection instanceof JedisCluster) {
                    luaResult =  (Long)((JedisCluster) nativeConnection).eval(unlockLua, Collections.singletonList(key), Collections.singletonList(uniqueId));
                }
                if(nativeConnection instanceof Jedis){
                    luaResult =  (Long)((Jedis)nativeConnection).eval(unlockLua, Collections.singletonList(key), Collections.singletonList(uniqueId));
                }
                return luaResult != 0;
            }
        });
    }
}

分布式锁接口类

public interface DistributeLock {
    void lock(String key,String uniqueId,Long expireTime);
    boolean tryLock(String key,String uniqueId,Long expireTime);
    void unlock(String key,String uniqueId);
}

业务回调类

public interface LockedCallback<R> {
    R callback();
}

redisTemplate配置类

@Bean("cardRedisTemplate")
    public <String,V> RedisTemplate<String,V> cardRedisTemplate(RedisConnectionFactory factory) {
        RedisTemplate<String,V> template = new RedisTemplate<>();
        template.setConnectionFactory(factory);

        //定义key序列化方式
        StringRedisSerializer stringRedisSerializer = new StringRedisSerializer();//Long类型会出现异常信息;需要我们上面的自定义key生成策略,一般没必要
        template.setKeySerializer(stringRedisSerializer);

        //定义value的序列化方式
        Jackson2JsonRedisSerializer jackson2JsonRedisSerializer = new Jackson2JsonRedisSerializer(Object.class);
        ObjectMapper om = new ObjectMapper();
        om.setVisibility(PropertyAccessor.ALL, JsonAutoDetect.Visibility.ANY);
        om.enableDefaultTyping(ObjectMapper.DefaultTyping.NON_FINAL);
        jackson2JsonRedisSerializer.setObjectMapper(om);
        template.setValueSerializer(jackson2JsonRedisSerializer);
        template.setHashValueSerializer(jackson2JsonRedisSerializer);

        template.afterPropertiesSet();
        return template;
    }

在你写分布式锁的时候最好先看下这篇文章redis分布式锁的正确姿势
考虑好这些问题之后再开始动手写。

猜你喜欢

转载自blog.csdn.net/mayongzhan_csdn/article/details/80748914