在一次产品需求设计中，有这样一种场景，对于一个工单，专员可以选择将工单挂起（一次挂起2小时；二次挂起12小时），或者转存（表单输入转存的天数），然后提交工单到既定时间后，该工单需要重新被领单，然后再次审核。对于工单，有优先级概念，意味着优先级越大，该工单则需要优先被派单，对于这种场景，我们可以基于Redis来解决场景问题。

一、业务概述

我们假定设置两个队列，一个队列维护正式工单，另一个队列维护挂起工单。对于挂起操作，我们通过Redis设置key有效时间，当key失效时，客户端监听失效事件，获取工单，实现挂起工单队列的移除，正式队列的入队即可。

业务流程图：

二、代码实现

Redis相关学习视频推荐：https://www.bilibili.com/video/BV1sT4y1j7G8

整个业务实现，可以分为三个大模块。

队列模块，通过WorkOrderQueueTransfer带外提供内部的调用。
工单模块、通过OperateStrategyManager管理器实现工单相关场景的操作。
事件监听、通过继承KeyExpirationEventMessageListener，实现对Key过期失效的监听（特别说明：Redis服务器redis.conf需要开启Key过期失效通知）。

2.1、工单队列实现

我们可以基于Redis ZSet数据存储，ZSet是个有序集合，可以实现基于score排序。

2.1.1、定义上下文类（WorkOrderContext）

/**
 * @description: 工单上下文对象
 * @Date : 2020/7/13 下午4:28
 * @Author : 石冬冬-Seig Heil
 */
@Data
@NoArgsConstructor
@AllArgsConstructor
@Builder
public class WorkOrderContext {
    /**
     * 是否测试人员专用
     */
    private boolean isTest;
    /**
     * 工单号
     */
    private WorkOrder worOrder;
    /**
     * 队列类型
     */
    private QueueType queueType;


    /**
     * 创建-正式队列（立即需要被派单
     * @return
     */
    public static WorkOrderContext buildImmediate() {
        return WorkOrderContext.builder().queueType(QueueType.immediate).build();
    }

    /**
     * 创建-挂起队列（挂起n小时执行）
     * @return
     */
    public static WorkOrderContext buildSuspended() {
        return WorkOrderContext.builder().queueType(QueueType.suspended).build();
    }


    /**
     * 转存队列（转存n天后执行）
     * @return
     */
    public static WorkOrderContext buildStored() {
        return WorkOrderContext.builder().queueType(QueueType.stored).build();
    }

    /**
     * 创建-正式队列（立即需要被派单）
     *
     * @param workCode
     * @param priority
     * @return
     */
    public static WorkOrderContext buildImmediate(String workCode, double priority) {
        WorkOrder workOrder = WorkOrder.builder().workCode(workCode).priority(priority).delayedTime(0).build();
        return WorkOrderContext.builder().worOrder(workOrder).queueType(QueueType.immediate).build();
    }

    /**
     * 创建-挂起队列（挂起n小时执行）
     *
     * @param workCode
     * @param priority
     * @param delayedTime
     * @return
     */
    public static WorkOrderContext buildSuspended(String workCode, double priority, long delayedTime) {
        WorkOrder workOrder = WorkOrder.builder().workCode(workCode).priority(priority).delayedTime(delayedTime).build();
        return WorkOrderContext.builder().worOrder(workOrder).queueType(QueueType.suspended).build();
    }

    /**
     * 转存队列（转存n天后执行）
     *
     * @param workCode
     * @param priority
     * @param delayedTime
     * @return
     */
    public static WorkOrderContext buildStored(String workCode, double priority, long delayedTime) {
        WorkOrder workOrder = WorkOrder.builder().workCode(workCode).priority(priority).delayedTime(delayedTime).build();
        return WorkOrderContext.builder().worOrder(workOrder).queueType(QueueType.stored).build();
    }


    /**
     * 队列类型
     */
    public enum QueueType {
        /**
         * 正式队列（立即需要被派单）
         */
        immediate,
        /**
         * 挂起队列（挂起n小时执行）
         */
        suspended,
        /**
         * 转存队列（转存n天后执行）
         */
        stored
    }

    @Data
    @NoArgsConstructor
    @AllArgsConstructor
    @Builder
    public static class WorkOrder {
        /**
         * 工单号
         */
        private String workCode;
        /**
         * 优先级
         */
        private double priority;
        /**
         * 延迟时间
         */
        private long delayedTime;
    }
}

2.1.2、定义抽象缓存类（AbstractCacheManager）

该抽象类定义了一个方法，抽象定义了一个具有重试机制的方法。其中注入的BeanRedisService是咱们子类基于Redis API的封装。

/**
 * @description: 抽象缓存管理器
 * @Date : 2020/7/18 下午9:41
 * @Author : 石冬冬-Seig Heil
 */
@Slf4j
public abstract class AbstractCacheManager {

    final int MAX_RETRIES = 3;

    @Autowired
    RedisService redisService;

    /**
     * 重试操作
     * @param retries 重试次数
     * @param context 上下文
     * @param call 重试动作
     */
    public <T> void retry(int retries, T context, Function<Integer,Boolean> call){
        boolean done = false;
        int retry = 1;
        do {
            try {
                done = call.apply(retry);
                log.info("[retry] context={},retry={},done={}", JSONObject.toJSON(context),retry,done);
                retry ++;
                TimeUnit.MILLISECONDS.sleep(100);
            } catch (Exception e) {
                log.error("[retry] 异常 ctx={}", JSONObject.toJSON(context),e);
                retry ++;
            }
        }while (retry <= retries && !done);
    }
}

2.1.3、定义基于Redis Cache的Manager类（WorkOrderCacheManager）

该类的主要作用，基于Redis String对象存储，实现具有Key失效机制的存储。

内部静态类CacheValue，作为Redis String对象存储的Value值。
内部枚举类CacheType，维护了缓存Key的业务前缀。
特别说明的是，我们构成Redis String存储Key的命名规则，例如：carCarthage:stored_cache_单号。

/**
 * @description: 工单缓存管理器
 * @Date : 2020/7/14 下午4:28
 * @Author : 石冬冬-Seig Heil
 */
@Component
@Slf4j
public class WorkOrderCacheManager extends AbstractCacheManager{

    /**
     * 设置缓存并设置缓存失效日期
     * @param cache
     */
    public void setCacheInExpire(CacheValue cache){
        retry(MAX_RETRIES,cache,idx -> {
            String redisKey = redisService.getKey(getRedisKeySuffix(cache.getType(),cache.getWorkCode()));
            redisService.set(redisKey, JSONObject.toJSONString(cache),cache.getExpireSeconds());
            log.info("[setCacheInExpire],redisKey={},CacheValue={}",redisKey,JSONObject.toJSONString(cache));
            return Boolean.TRUE;
        });
    }

    /**
     * 查询某个工单号的缓存值
     * @param cacheType 缓存类型 {@link CacheType}
     * @param workCode 工单号
     * @return
     */
    public CacheValue get(CacheType cacheType,String workCode){
        String redisKey = redisService.getKey(getRedisKeySuffix(cacheType,workCode));
        String value =  redisService.get(redisKey,String.class);
        return JSONObject.parseObject(value,CacheValue.class);
    }

    /**
     * 从上下文队列类型获取队列redis key
     * @param cacheType 缓存类型 {@link CacheType}
     * @param workCode 工单号
     * @return
     */
    String getRedisKeySuffix(CacheType cacheType,String workCode){
        switch (cacheType){
            case stored_cache:
                return CacheType.stored_cache.getKey() + workCode;
            case suspended_cache:
                return CacheType.suspended_cache.getKey() + workCode;
            default:
                break;
        }
        return null;
    }

    /**
     * 缓存值
     */
    @Data
    @NoArgsConstructor
    @AllArgsConstructor
    @Builder
    public static class CacheValue{
        /**
         * 缓存类型
         */
        private CacheType type;
        /**
         * 工单号
         */
        private String workCode;
        /**
         * 优先级
         */
        private double priority;
        /**
         * 延迟截止时间(单位：时间戳)
         */
        private long delayedTime;
        /**
         * 缓存失效时间(单位：秒)
         */
        private long expireSeconds;

        /**
         * 创建-挂起队列（挂起n小时执行）
         *
         * @param workCode
         * @param priority
         * @param delayedTime
         * @param expireSeconds
         * @return
         */
        public static CacheValue buildSuspended(String workCode, double priority, long delayedTime,long expireSeconds) {
            return CacheValue.builder()
                    .type(CacheType.suspended_cache)
                    .workCode(workCode)
                    .priority(priority)
                    .delayedTime(delayedTime)
                    .expireSeconds(expireSeconds)
                    .build();
        }

        /**
         * 转存队列（转存n天后执行）
         *
         * @param workCode
         * @param priority
         * @param delayedTime
         * @param expireSeconds
         * @return
         */
        public static CacheValue buildStored(String workCode, double priority, long delayedTime,long expireSeconds) {
            return CacheValue.builder()
                    .type(CacheType.stored_cache)
                    .workCode(workCode)
                    .priority(priority)
                    .delayedTime(delayedTime)
                    .expireSeconds(expireSeconds)
                    .build();
        }
    }

    /**
     * 实现 挂起|转存 缓存key
     */
    @Getter
    public enum CacheType{
        stored_cache("stored_cache_"),
        suspended_cache("suspended_cache_"),
        ;
        CacheType(String key) {
            this.key = key;
        }
        private String key;
    }
}

2.1.4、工单队列管理器（WorkOrderQueueManager）

该类基于Redis ZSet 对象结构的有序集合，可以实现按照优先级出队。

相关方法介绍：

String getRedisKey(WorkOrderContext context)：从 WorkOrderContext中获取队列类型，返回集合要存储的Key。
Long queueSize(WorkOrderContext context)：返回队列大小
Boolean leftPush(WorkOrderContext context)：执行处理（入队操作）
Boolean leftPushIfAbsent(WorkOrderContext context)：执行处理（入队操作），如果入队元素缺席则入队，返回true；否则返回false。
Long remove(WorkOrderContext context)：从队列移除某个元素
WorkOrderContext.WorkOrder pop(WorkOrderContext context)：从集合中获取评分最小的成员出队
Set<WorkOrderContext.WorkOrder> rank(WorkOrderContext context)：按照升序查看队列中所有成员
Long removeRange(String key, long start, long end)：按照范围移除队列元素
Long removeValues(String key, List<Object> values)：移除指定元素
long getDelayedTime(WorkOrderContext.QueueType queueType,String workCode)：获取对应工单的延迟时间（适用于挂起和转存）

/**
 * @description: 工单队列管理器
 * @Date : 2020/7/14 下午4:28
 * @Author : 石冬冬-Seig Heil
 */
@Component
@Slf4j
public class WorkOrderQueueManager extends AbstractCacheManager{

    final String LOCK_KEY = "ZSET_ATOMIC_LOCK";

    @Autowired
    ZSetOperations<String, Object> zSetOperations;

    @Autowired
    WorkOrderCacheManager workOrderCacheManager;

    /**
     * 从上下文队列类型获取队列redis key
     * @param context
     * @return
     */
    String getRedisKey(WorkOrderContext context){
        String keySuffix = null;
        switch (context.getQueueType()){
            case immediate:
                keySuffix = CarthageConst.WorkOrderKey.IMMEDIATE_QUEUE_DEFAULT;
                break;
            case stored:
                keySuffix = CarthageConst.WorkOrderKey.STORED_QUEUE_DEFAULT;
                break;
            case suspended:
                keySuffix = CarthageConst.WorkOrderKey.SUSPENDED_QUEUE_DEFAULT;
                break;
            default:
                break;
        }
        if(null != keySuffix){
            if(context.isTest()){
                keySuffix += CarthageConst.TEST_SUFFIX;
            }
            return redisService.getKey(keySuffix);
        }
        return null;
    }

    /**
     * 返回队列大小
     * @param context
     * @return
     */
    public Long queueSize(WorkOrderContext context) {
        return zSetOperations.size(getRedisKey(context));
    }

    /**
     * 执行处理（入队操作）
     * @param context
     * @return
     */
    public Boolean leftPush(WorkOrderContext context) {
        String redisKey = getRedisKey(context);
        String workCode = context.getWorOrder().getWorkCode();
        double priority = context.getWorOrder().getPriority();
        Boolean action = zSetOperations.add(redisKey,workCode,priority);
        if(Objects.equals(Boolean.FALSE,action)){
            Long value = zSetOperations.rank(redisKey,workCode);
            log.info("[Queue.leftPush],hasLeftPushed,action={},value={}, context={}", action,value,JSONObject.toJSON(context));
            if(Objects.nonNull(value)){
                return Boolean.TRUE;
            }
        }
        log.info("[Queue.leftPush] context={}", JSONObject.toJSON(context));
        retry(MAX_RETRIES,context,idx -> action);
        return Optional.ofNullable(action).orElse(Boolean.FALSE);
    }

    /**
     * 执行处理（入队操作）
     * 如果入队元素缺席则入队，返回true；否则返回false。
     * @param context
     * @return
     */
    public Boolean leftPushIfAbsent(WorkOrderContext context) {
        String redisKey = getRedisKey(context);
        String workCode = context.getWorOrder().getWorkCode();
        double priority = context.getWorOrder().getPriority();
        Boolean action = zSetOperations.add(redisKey,workCode,priority);
        log.info("[WorkOrderQueue.leftPushIfAbsent,action={},context={}",action, JSONObject.toJSON(context));
        return Optional.ofNullable(action).orElse(Boolean.FALSE);
    }

    /**
     * 从队列移除某个元素
     * @param context
     * @return
     */
    public Long remove(WorkOrderContext context){
        String redisKey = getRedisKey(context);
        String workCode = context.getWorOrder().getWorkCode();
        log.info("[WorkOrderQueue.remove] context={}", JSONObject.toJSON(context));
        Long rem = zSetOperations.remove(redisKey,workCode);
        Long action = Optional.ofNullable(rem).orElse(0L);
        retry(MAX_RETRIES,context,idx -> action.longValue() > 0);
        return action;
    }

    /**
     * 从集合中获取评分最小的成员出队
     * @param context
     * @return
     */
    public WorkOrderContext.WorkOrder pop(WorkOrderContext context) {
        WorkOrderContext.WorkOrder workOrder = null;
        try {
            String redisKey = getRedisKey(context);
            //通过分布式锁，实现 zset 的 zpopmin 命令操作
            boolean locked = redisService.lock(LOCK_KEY,5000);
            if(locked){
                //1、取出第一个最小评分元素
                Set<ZSetOperations.TypedTuple<Object>> set = redisService.zSetOperations().rangeWithScores(redisKey,0,0);
                if(set.isEmpty()){
                    return null;
                }
                //2、移除该最小评分元素
                Long value = redisService.zSetOperations().removeRange(redisKey,0,0);
                retry(MAX_RETRIES,context,idx -> value.longValue() > 0);
                //3、返回出队成员
                workOrder = WorkOrderContext.WorkOrder.builder().build();
                for(ZSetOperations.TypedTuple<Object> each : set){
                    workOrder.setWorkCode(each.getValue().toString());
                    workOrder.setPriority(each.getScore());
                    workOrder.setDelayedTime(0);
                    break;
                }
            }
        } catch (Exception e) {
            log.error("[WorkOrderQueue.pop] exception ctx={}", JSONObject.toJSON(context));
        }finally {
            redisService.unlock(LOCK_KEY);
        }
        return workOrder;
    }

    /**
     * 按照升序查看队列中所有成员
     * @param context
     * @return
     */
    public Set<WorkOrderContext.WorkOrder> rank(WorkOrderContext context) {

        Set<ZSetOperations.TypedTuple<Object>> set = redisService.zSetOperations().rangeWithScores(getRedisKey(context),0,-1);

        Set<WorkOrderContext.WorkOrder> members = Sets.newLinkedHashSetWithExpectedSize(set.size());

        set.forEach(each -> {
            WorkOrderContext.WorkOrder every = WorkOrderContext.WorkOrder.builder()
                    .workCode(each.getValue().toString())
                    .priority(each.getScore())
                    .delayedTime(getDelayedTime(context.getQueueType(),each.getValue().toString()))
                    .build();
            members.add(every);
        });
        return members;
    }

    /**
     * 按照范围移除队列元素
     * @param key
     * @param start
     * @param end
     * @return
     */
    public Long removeRange(String key, long start, long end){
        String redisKey = redisService.getKey(key);
        Long count = zSetOperations.removeRange(redisKey,start,end);
        log.info("[WorkOrderQueue.removeRange] redisKey={},start={},end={},count={}", redisKey,start,end,count);
        return count;
    }

    /**
     * 移除指定元素
     * @param key
     * @param values
     * @return
     */
    public Long removeValues(String key, List<Object> values){
        String redisKey = redisService.getKey(key);
        LongAdder longAdder = new LongAdder();
        values.forEach(each -> {
            Long count = zSetOperations.remove(redisKey,each);
            longAdder.add(count);
        });
        Long count =  longAdder.longValue();
        log.info("[WorkOrderQueue.removeValues] redisKey={},values={},count={}", redisKey,JSONObject.toJSONString(values),count);
        return count;
    }

    /**
     * 获取对应工单的延迟时间（适用于 挂起 和 转存）
     * @param queueType
     * @param workCode
     * @return
     */
    long getDelayedTime(WorkOrderContext.QueueType queueType,String workCode){
        long delayedTime = 0;
        WorkOrderCacheManager.CacheType cacheType = null;
        switch (queueType){
            case suspended:
                cacheType = WorkOrderCacheManager.CacheType.suspended_cache;
                break;
            case stored:
                cacheType = WorkOrderCacheManager.CacheType.stored_cache;
                break;
            default:
                break;
        }
        if(null != cacheType){
            WorkOrderCacheManager.CacheValue cacheValue = workOrderCacheManager.get(cacheType,workCode);
            if(null != cacheValue){
                delayedTime = cacheValue.getDelayedTime();
            }
        }
        return delayedTime;
    }
}

2.2、工单队列转移管理器

2.2.1、工单队列转移管理器（WorkOrderQueueTransfer）

该类实现延迟队列出队、正式队列入队的原子操作（通过Redis分布式锁实现）。

/**
 * @description: 工单队列转移管理器
 * @Date : 2020/7/23 下午6:15
 * @Author : 石冬冬-Seig Heil
 */
@Component
@Slf4j
public class WorkOrderQueueTransfer extends AbstractCacheManager{

    final static String ATOMIC_KEY = "delayed_queue_key_expire_lock_{0}";

    final static long ATOMIC_KEY_EXPIRE = 5000;

    @Autowired
    RedisService redisService;

    @Autowired
    WorkOrderQueueManager workOrderQueueManager;

    @Autowired
    WorkOrderCacheManager workOrderCacheManager;

    /**
     * 从[挂起|暂存]队列转移到正式队列中
     * @param cacheType 挂起|暂存
     * @param delayedContext
     * @return
     */
    public Boolean transferImmediateQueue(WorkOrderCacheManager.CacheType cacheType,WorkOrderContext delayedContext){
        boolean tryLock = false;
        Boolean done = Boolean.FALSE;
        String lockKey = null;
        try {
            WorkOrderContext.WorkOrder workOrder = delayedContext.getWorOrder();
            lockKey = redisService.getKey(MessageFormat.format(ATOMIC_KEY,workOrder.getWorkCode()));
            tryLock = redisService.lock(lockKey,ATOMIC_KEY_EXPIRE);
            if(tryLock){
                //1、构建正式队列
                WorkOrderContext immediateContext = WorkOrderContext.buildImmediate(workOrder.getWorkCode(),workOrder.getPriority());
                done = workOrderQueueManager.leftPushIfAbsent(immediateContext);
                //2、从当前延迟队列移除该元素
                Long count = workOrderQueueManager.remove(delayedContext);
                log.info("[挂起|转存队remove],count={},delayedContext={}", count,JSONObject.toJSONString(delayedContext));
            }
        } catch (Exception e) {
            log.error("[transferImmediateQueue]异常,delayedContext={},cacheType={}", JSONObject.toJSONString(delayedContext),cacheType);
        }finally {
            if(Objects.nonNull(lockKey) && tryLock){
                redisService.unlock(lockKey);
            }
        }
        return Optional.ofNullable(done).orElse(Boolean.FALSE);
    }
}

2.3、Redis过期key监听

2.3.1、Redis过期回调监听（RedisKeyExpirationListener）

/**
 * @description: Redis过期回调监听
 * @Date : 2020/7/18 上午10:43
 * @Author : 石冬冬-Seig Heil
 */
@Component
@Slf4j
public class RedisKeyExpirationListener extends KeyExpirationEventMessageListener {

    final static String STORED_CACHE_KEY_PREFIX = WorkOrderCacheManager.CacheType.stored_cache.getKey();

    final static String SUSPENDED_CACHE_KEY_PREFIX = WorkOrderCacheManager.CacheType.suspended_cache.getKey();

    @Autowired
    TraceLogService traceLogService;

    @Autowired
    RedisService redisService;

    @Autowired
    WorkOrderService workOrderService;

    @Autowired
    DelayedScheduledOperateBridge delayedScheduledOperateBridge;

    public RedisKeyExpirationListener(RedisMessageListenerContainer listenerContainer) {
        super(listenerContainer);
    }

    @Override
    public void onMessage(Message message, byte[] pattern) {
        Date startTime = TimeTools.createNowTime();
        String expiredKey = message.toString();
        String bizPrefix = redisService.getKeyPrefix().getName();
        if(!expiredKey.startsWith(bizPrefix)){
            return;
        }
        String caseOfStored = redisService.getKey(STORED_CACHE_KEY_PREFIX);
        String caseOfSuspended = redisService.getKey(SUSPENDED_CACHE_KEY_PREFIX);

        WorkOrderCacheManager.CacheType cacheType;
        WorkOrderContext.QueueType queueType;
        if(expiredKey.startsWith(caseOfStored)){
            queueType = WorkOrderContext.QueueType.stored;
            cacheType = WorkOrderCacheManager.CacheType.stored_cache;
        }else if(expiredKey.startsWith(caseOfSuspended)){
            queueType = WorkOrderContext.QueueType.suspended;
            cacheType = WorkOrderCacheManager.CacheType.suspended_cache;
        }else{
            return;
        }
        String workCode = getWorkCode(expiredKey);
        log.info("监听到 redis key=[{}] 已过期",expiredKey);
        if(Objects.nonNull(workCode)){
            log.info("监听到 redis key=[{}],挂起|转存工单开始处理，workCode={}",expiredKey,workCode);
            WorkOrder workOrder = workOrderService.queryOne(workCode);
            if(Objects.isNull(workOrder)){
                log.info("监听到 redis key=[{}],挂起|转存工单开始处理，未找到工单，workCode={}",expiredKey,workCode);
                return;
            }
            WorkOrderContext delayedContext = WorkOrderContext.builder()
                    .worOrder(WorkOrderContext.WorkOrder.builder().delayedTime(5).priority(workOrder.getCasePriority()).workCode(workOrder.getWorkCode()).build())
                    .queueType(queueType). build();
            Boolean done = delayedScheduledOperateBridge.transferImmediateQueue(cacheType,delayedContext);
            saveTraceLog(delayedContext,done,traceLog -> {
                JSONObject requestBody = new JSONObject();
                requestBody.put("expiredKey",expiredKey);
                requestBody.put("workCode",workCode);
                traceLog.setRequestBody(requestBody.toJSONString());
                traceLog.setRequestTime(startTime);
            });
        }
    }

    /**
     * traceLog入库
     * @param context
     * @param done
     * @param consumer
     */
    void saveTraceLog(WorkOrderContext context, Boolean done, Consumer<TraceLog> consumer){
        try {
            String hostAddress = InetAddress.getLocalHost().getHostAddress();
            JSONObject responseBody = new JSONObject();
            responseBody.put("workOrderContext",context);
            responseBody.put("transferImmediateQueue",done);
            TraceLog traceLog = TraceLog.builder()
                    .appCode(context.getWorOrder().getWorkCode())
                    .url("["+hostAddress+"]redisKeyExpirationListener.onMessage")
                    .target(this.getClass().getPackage().getName() + "." + this.getClass().getSimpleName())
                    .responseBody(responseBody.toJSONString())
                    .responseTime(TimeTools.createNowTime())
                    .traceType(TraceTypeEnum.REDIS_KEY_EXPIRE.getIndex())
                    .build();
            consumer.accept(traceLog);
            traceLogService.insertRecord(traceLog);
        } catch (Exception e) {
            log.error("saveTraceLog exception,[context={}]",JSONObject.toJSONString(context),e);
        }
    }

    /**
     * 从字符串截取制定的工单号
     * @param value
     * @return
     */
    String getWorkCode(String value){
        return value.substring(value.lastIndexOf("_") + 1);
    }
}

2.4、延迟订单既定处理桥接器

该类的主要作用，就是通过WorkOrderQueueTransfer实现队列元素的转移，同时通过OperateStrategyManager实现工单的数据库表操作。

/**
 * @description: 延迟订单既定处理处理桥接
 * 场景描述：挂起|转存 工单到既定时间的处理
 * @Date : 2020/7/23 22:20
 * @Author : Seig Heil
 */
@Slf4j
@Component
public class DelayedScheduledOperateBridge {

    static final String LOCK_KEY = CarthageConst.KEY_EXPIRE_LISTENER_LOCK;

    static final long EXPIRE_SECONDS = 120;

    @Autowired
    RedisService redisService;

    @Autowired
    WorkOrderQueueTransfer workOrderQueueTransfer;

    @Autowired
    OperateStrategyManager operateStrategyManager;

    /**
     * 实现从延迟队列到正式队列的转移业务处理，同时更新工单的状态
     * @param cacheType
     * @param delayedContext
     * @return
     */
    public Boolean transferImmediateQueue(WorkOrderCacheManager.CacheType cacheType, WorkOrderContext delayedContext){
        String workCode = delayedContext.getWorOrder().getWorkCode();
        boolean tryLock = false;
        String redisKey = null;
        try {
            redisKey = redisService.getKey(MessageFormat.format(LOCK_KEY,workCode));
            tryLock = redisService.lock(redisKey,EXPIRE_SECONDS);
            if(!tryLock){
                log.info("[DelayedScheduledOperateBridge.tryLock={}获取锁失败,redisKey={}]挂起|转存既定时间处理幂等，workCode={}",tryLock,redisKey,workCode);
            }
            if(tryLock){
                log.info("[DelayedScheduledOperateBridge.tryLock={}获取锁成功,redisKey={}]挂起|转存既定时间处理幂等，workCode={}",tryLock,redisKey,workCode);
                Boolean done = workOrderQueueTransfer.transferImmediateQueue(cacheType,delayedContext);
                if(!done.booleanValue()){
                    return Boolean.FALSE;
                }
                OperateContext operateContext = OperateContext.builder()
                        .operateStrategyEnum(OperateContext.OperateStrategyEnum.DELAYED_SCHEDULED_ORDER)
                        .operateParam( OperateContext.OperateParam.builder().workCode(workCode).build()
                        ).build();
                operateStrategyManager.execute(operateContext);
                log.info("[DelayedScheduledOperateBridge.transferImmediateQueue],delayedContext={},callResult={}",
                        JSONObject.toJSONString(delayedContext),JSONObject.toJSONString(operateContext.getExecuteResult()));
                return operateContext.getExecuteResult().isSuccess();
            }
        } catch (Exception e) {
            log.error("[DelayedScheduledOperateBridge]挂起|转存既定时间处理异常，workCode={},delayedContext={}",workCode,JSONObject.toJSONString(delayedContext));
        }finally {
            if(tryLock){
                redisService.unlock(redisKey);
            }
        }
        return false;
    }
}

2.5、工单操作管理器

该类的主要作用，就是对外暴露工单操作策略类的管理，外部无需关注策略类的存在，策略类实例的创建由该类负责。

通过OPERATE_STRATEGY_MAP维护枚举和策略类Bean的映射。
通过init()实现OPERATE_STRATEGY_MAP容器的初始化工作。
通过Result<String> execute(OperateContext context)实现对外提供策略类的操作。

/**
 * @description: GPS工单操作策略管理类
 * @Date : 2020/7/15 下午5:43
 * @Author : 石冬冬-Seig Heil
 */
@Component
@Slf4j
public class OperateStrategyManager {

    static final Map<OperateContext.OperateStrategyEnum, AbstractOperateStrategy> OPERATE_STRATEGY_MAP = Maps.newHashMapWithExpectedSize(6);

    @Autowired
    CreateOperateStrategy createOperateStrategy;

    @Autowired
    AllotOrderOperateStrategy allotOrderOperateStrategy;

    @Autowired
    SubmitWithFinishOperateStrategy submitWithFinishOperateStrategy;

    @Autowired
    SubmitWithStoreOperateStrategy submitWithStoreOperateStrategy;

    @Autowired
    SubmitWithSuspendOperateStrategy submitWithSuspendOperateStrategy;

    @Autowired
    DelayedScheduledOperateStrategy delayedScheduledOperateStrategy;

    @Autowired
    AssignOrderOperateStrategy assignOrderOperateStrategy;

    @PostConstruct
    private void init() {
        OPERATE_STRATEGY_MAP.put(OperateContext.OperateStrategyEnum.INIT_WORK_ORDER, createOperateStrategy);
        OPERATE_STRATEGY_MAP.put(OperateContext.OperateStrategyEnum.ALLOT_WORK_ORDER, allotOrderOperateStrategy);
        OPERATE_STRATEGY_MAP.put(OperateContext.OperateStrategyEnum.STORE_WORK_ORDER, submitWithStoreOperateStrategy);
        OPERATE_STRATEGY_MAP.put(OperateContext.OperateStrategyEnum.SUSPEND_WORK_ORDER, submitWithSuspendOperateStrategy);
        OPERATE_STRATEGY_MAP.put(OperateContext.OperateStrategyEnum.FINISH_WORK_ORDER, submitWithFinishOperateStrategy);
        OPERATE_STRATEGY_MAP.put(OperateContext.OperateStrategyEnum.DELAYED_SCHEDULED_ORDER, delayedScheduledOperateStrategy);
        OPERATE_STRATEGY_MAP.put(OperateContext.OperateStrategyEnum.ASSIGN_ORDER, assignOrderOperateStrategy);
    }

    /**
     * 对外提供对策略类的调用
     * @param context
     * @return
     */
    public Result<String> execute(OperateContext context) {
        StopWatch stopWatch = new StopWatch();
        stopWatch.start("OperateStrategyManager.execute");
        AbstractOperateStrategy operateStrategy = OPERATE_STRATEGY_MAP.get(context.getOperateStrategyEnum());
        context.buildExecuteResultWithSuccess();
        operateStrategy.execute(context);
        Result<Boolean> executeResult = context.getExecuteResult();
        if(context.getExecuteResult().isSuccess()) {
            return Result.suc(executeResult.getMsg());
        }
        stopWatch.stop();
        long spendMillSeconds = stopWatch.getLastTaskTimeMillis();
        long duration = (System.currentTimeMillis() - spendMillSeconds) / 1000;
        String executeResultMsg = executeResult.getMsg();
        log.info("[execute] done,duration={},executeResultMsg={}",duration,executeResultMsg);
        return Result.fail(RemoteEnum.FAILURE, executeResultMsg);
    }
}

2.6、工单策略类实现

由于工单涉及到创建、挂起、转存、处理完结等操作，因此对于这类场景我们可以通过策略类来实现。

2.6.1、挂起操作

/**
 * @description: 提交调查结果（挂起操作）-策略类
 * @Date : 2020/7/15 下午5:32
 * @Author : 石冬冬-Seig Heil
 */
@Slf4j
@Component
public class SubmitWithSuspendOperateStrategy extends AbstractSubmitOperateStrategy{

    static final Map<MoveToEnum,AttentionEventEnum> suspend_to_attention_event_map = new HashMap<>();

    static final Map<MoveToEnum,WorkOrderStatusEnum.SubStatusEnum> suspend_to_sub_status_map = new HashMap<>();

    static final Map<MoveToEnum,Integer> suspend_count_map = new HashMap<>();

    static {

        suspend_to_attention_event_map.put(MoveToEnum.SUSPENDED_AT_ONCE,AttentionEventEnum.SUSPENDED_AT_ONCE);
        suspend_to_attention_event_map.put(MoveToEnum.SUSPENDED_AT_TWICE,AttentionEventEnum.SUSPENDED_AT_TWICE);

        suspend_to_sub_status_map.put(MoveToEnum.SUSPENDED_AT_ONCE,WorkOrderStatusEnum.SubStatusEnum.SUSPENDED_AT_ONCE);
        suspend_to_sub_status_map.put(MoveToEnum.SUSPENDED_AT_TWICE,WorkOrderStatusEnum.SubStatusEnum.SUSPENDED_AT_TWICE);

        suspend_count_map.put(MoveToEnum.SUSPENDED_AT_ONCE,1);
        suspend_count_map.put(MoveToEnum.SUSPENDED_AT_TWICE,2);

        log.info("init... suspend_to_attention_event_map={}",suspend_to_attention_event_map.toString());
        log.info("init... suspend_to_sub_status_map={}",suspend_to_sub_status_map.toString());
        log.info("init... suspend_count_map={}",suspend_count_map.toString());
    }

    @Autowired
    DiamondConfigProxy diamondConfigProxy;

    @Override
    public void prepare(OperateContext context) {
        super.prepare(context);
        SurveyResult surveyResult = context.getSurveyResult();
        MoveToEnum moveToEnum = MoveToEnum.getByIndex(surveyResult.getMoveTo());
        AttentionEvent attentionEvent = suspend_to_attention_event_map.getOrDefault(moveToEnum,null);
        ATTENTION_EVENT_CONTEXT.set(attentionEvent);
        context.setAttentionEvent(attentionEvent);
    }

    @Override
    WorkOrder buildWorkOrder(OperateContext context){
        SurveyResult surveyResult = context.getSurveyResult();
        MoveToEnum moveToEnum = MoveToEnum.getByIndex(surveyResult.getMoveTo());
        WorkOrder workOrder = super.buildWorkOrder(context);
        workOrder.setSuspendedCount(suspend_count_map.getOrDefault(moveToEnum,0).intValue());
        workOrder.setMainStatus(WorkOrderStatusEnum.WAITING.getIndex());
        workOrder.setSubStatus(suspend_to_sub_status_map.get(moveToEnum).getIndex());
        workOrder.setIsFinished(Const.NON_INDEX);
        workOrder.setIsStore(Const.NON_INDEX);
        workOrder.setDelayedTime(context.getOperateParam().getDelayedTime());
        return workOrder;
    }

    @Override
    void operationExtend(OperateContext context) {
        long delayedTime = context.getOperateParam().getDelayedTime().getTime();
        int delayedSeconds = context.getOperateParam().getDelayedSeconds();
        WorkOrder workOrder = context.getWorkOrder();
        WorkOrderContext cxt = WorkOrderContext.buildSuspended(workOrder.getWorkCode(),workOrder.getCasePriority(),delayedTime);
        workOrderQueueManager.leftPush(cxt);

        WorkOrderCacheManager.CacheValue cacheValue = WorkOrderCacheManager.CacheValue.
                buildSuspended(workOrder.getWorkCode(),workOrder.getCasePriority(),delayedTime,delayedSeconds);
        workOrderCacheManager.setCacheInExpire(cacheValue);
        super.operationExtend(context);
    }

    @Override
    public void setDelayedTime(OperateContext context) {
        SurveyResult surveyResult = context.getSurveyResult();
        MoveToEnum moveToEnum = MoveToEnum.getByIndex(surveyResult.getMoveTo());
        DiamondConfig.SuspendOrderConfig suspendOrderConfig = diamondConfigProxy.suspendOrderConfig();
        Date delayedTime = TimeTools.createNowTime();
        int timeUnit = Calendar.HOUR_OF_DAY;
        int delayedSeconds = 0;
        int value = suspendOrderConfig.getConfig().getOrDefault(moveToEnum.name(),0);
        switch (suspendOrderConfig.getTimeUnit()){
            case "DAY":
                timeUnit = Calendar.DAY_OF_YEAR;
                delayedSeconds = value * 24 * 3600;
                break;
            case "HOUR":
                timeUnit = Calendar.HOUR_OF_DAY;
                delayedSeconds = value * 3600;
                break;
            case "MINUTE":
                timeUnit = Calendar.MINUTE;
                delayedSeconds = value * 60;
                break;
            case "SECOND":
                timeUnit = Calendar.SECOND;
                delayedSeconds = value;
                break;
            default:
                break;
        }

        TimeTools.addTimeField(delayedTime, timeUnit,value);
        context.getOperateParam().setDelayedTime(delayedTime);
        context.getOperateParam().setDelayedSeconds(delayedSeconds);
    }
}

2.6.2、转存操作

/**
 * @description: 提交调查结果（转存操作）-策略类
 * @Date : 2020/7/15 下午5:32
 * @Author : 石冬冬-Seig Heil
 */
@Slf4j
@Component
public class SubmitWithStoreOperateStrategy extends AbstractSubmitOperateStrategy{
    /**
     * 转存天数 换算 秒数
     */
    static final int DAY_TO_SECONDS = 24 * 60 * 60;

    @Override
    public void prepare(OperateContext context) {
        ATTENTION_EVENT_CONTEXT.set(AttentionEventEnum.STORE_ORDER);
        context.setAttentionEvent(AttentionEventEnum.STORE_ORDER);
        super.prepare(context);
    }

    @Override
    public boolean paramCheck(OperateContext context) {
        if(Objects.isNull(context.getSurveyResult().getDelayedDays())){
            context.buildExecuteResultWithFailure("[surveyResult.delayedDays]为空！");
        }
        if(context.getSurveyResult().getDelayedDays() == 0){
            context.buildExecuteResultWithFailure("等待天数[delayedDays]必须大于0！");
        }
        return super.paramCheck(context);
    }

    @Override
    WorkOrder buildWorkOrder(OperateContext context){
        WorkOrder workOrder = super.buildWorkOrder(context);
        workOrder.setMainStatus(WorkOrderStatusEnum.PENDING.getIndex());
        workOrder.setSubStatus(WorkOrderStatusEnum.SubStatusEnum.STORED.getIndex());
        workOrder.setIsFinished(Const.NON_INDEX);
        workOrder.setIsStore(Const.YES_INDEX);
        //setSuspendedCount 这里需要重置为0，转存后派单流程状态依赖该字段
        workOrder.setSuspendedCount(0);
        workOrder.setDelayedTime(context.getOperateParam().getDelayedTime());
        return workOrder;
    }

    @Override
    void operationExtend(OperateContext context) {
        long delayedTime = context.getOperateParam().getDelayedTime().getTime();
        int delayedSeconds = context.getOperateParam().getDelayedSeconds();
        WorkOrder workOrder = context.getWorkOrder();
        WorkOrderContext cxt = WorkOrderContext.buildStored(workOrder.getWorkCode(),workOrder.getCasePriority(),delayedTime);
        workOrderQueueManager.leftPush(cxt);

        WorkOrderCacheManager.CacheValue cacheValue = WorkOrderCacheManager.CacheValue.
                buildStored(workOrder.getWorkCode(),workOrder.getCasePriority(),delayedTime,delayedSeconds);
        workOrderCacheManager.setCacheInExpire(cacheValue);

        super.operationExtend(context);
    }

    @Override
    public void setDelayedTime(OperateContext context) {
        int delayedDays = context.getSurveyResult().getDelayedDays();
        Date delayedTime = TimeTools.createNowTime();
        TimeTools.addTimeField(delayedTime, Calendar.DAY_OF_YEAR,delayedDays);
        context.getOperateParam().setDelayedTime(delayedTime);
        context.getOperateParam().setDelayedSeconds(delayedDays * DAY_TO_SECONDS);
    }
}

三、交互UI介绍

系统链路日志

我们可以实现对于key失效事件监听，把工单落库，便于后期排查问题，无需再查看服务器日志。

队列监控

输出队列元素，直观看到当前挂起、转存的工单。我们可以直观看到key的剩余存活时间。

参数配置

挂起配置，尽管业务实现按照天维度、但是我代码依然实现了可以支持天、时、分、秒维度，提供了可扩展性。具体可以看 SubmitWithStoreOperateStrategy类方法void setDelayedTime(OperateContext context) 的实现。

四、总结

在接到该需求后，尽管这个只是其中需求的一小部分，整个产品需求33页，内容还是丰富的，对于延迟队列操作，我在调研技术方案时，也考虑过其他方法，譬如java自带的带有延迟特性的队列DelayedQueue（对于分布式多实例场景它就不适合了），以及可以通过RabbitMQ实现（感觉实现比较复杂），最终选择了Redis（可以利用相关数据特性比如ZSet,String,Expire），技术实现比较简单上手。

基于Redis解决业务场景中延迟队列就这么简单！

一、业务概述

二、代码实现

三、交互UI介绍

四、总结