redis队列及多线程应用
由于xxx平台上自己的博客已经很久没更新了，一直以来都是用的印象笔记来做工作中知识的积累存根，不知不觉印象笔记里已经有了四、五百遍文章。为了从新开始能与广大攻城狮共同提高技术能力与水平，随决心另起炉灶在新的博客与大家分享
经过一段时间项目的沉淀之后，对实际应用中的多线程开发及队列使用产生了深厚的兴趣，也将<<java并发编程实战>>仔细的阅读了两三遍，也看了很多并发编程的实践项目，也有了深刻的理解与在实践中合理应用队列、多线程开发的应用场景
1、真实应用场景描述：
由于一段时间以来要针对公司整个电商平台包括官网、移动端所有的交易数据进行统计，统计指标包括：pv、uv、实付金额、转化率、毛利率等等，按照各种不同的维度来统计计算出当前交易系统的各个指标的数据，但要求该项目是独立的，没有任务其它资源的协助及接品提供。经过一番xxxx思考讨论之后。业务上决定用以下解决方案：
A: 用一个定时服务每隔10秒去别的系统数据库抓取上一次查询时间以来新确认的订单（这种订单表示已经支付完在或者客户已经审核确认了），然后将这些订单的唯一编号放入redis队列。
B: 由于用到了队列，根据经验自然而然的想到了 启动单独的线程去redis队列中不断获取要统计处理的订单编号，然后将获取到的订单编号放入线程池中进行订单的统计任务处理。
开发实现：
FetchConfirmOrdersFromErpJob.java
1 /** 2 * 1、从redis中获取上次查询的时间戳 3 * 2、将当前时间戳放入到redis中，以便 下次按这个时间查询 4 * 3、去erp订单表查询confirm_time>=上次查询的时间的订单，放入队列中 5 */ 6 @Scheduled(cron = "0/30 * * * * ?") 7 public void start(){ 8 logger.info("FetchConfirmOrdersFromErpJob start................."+ new Date()); 9 StopWatch watch=new StopWatch(); 10 watch.start(); 11 //上次查询的时间 12 String preQueryTimeStr=this.readRedisService.get(Constans.CACHE_PREQUERYORDERTIME); 13 14 Date now=new Date(); 15 if(StringUtils.isBlank(preQueryTimeStr)){ 16 preQueryTimeStr=DateFormatUtils.format(DateUtils.addHours(now, -1), Constans.DATEFORMAT_PATTERN_YYYYMMDDHHMMSS);//第一次查询之前一个小时的订单 17 // preQueryTimeStr="2015-05-07 10:00:00";//本地测试的时候使用 18 } 19 //设置当前时间为上次查询的时间 20 this.writeRedisService.set(Constans.CACHE_PREQUERYORDERTIME, DateFormatUtils.format(now, Constans.DATEFORMAT_PATTERN_YYYYMMDDHHMMSS)); 21 22 List<Map<String, Object>> confirmOrderIds = this.erpOrderService.selectOrderIdbyConfirmtime(preQueryTimeStr); 23 if(confirmOrderIds==null){ 24 logger.info("query confirmOrderIds is null,without order data need dealth.........."); 25 return; 26 } 27 for (Map<String, Object> map : confirmOrderIds) {
//将订单编号放入队列中 28 this.writeRedisService.lpush(Constans.CACHE_ORDERIDS, map.get("channel_orderid").toString()); 29 logger.info("=======lpush orderid:"+map.get("channel_orderid").toString()); 30 } 31 32 watch.stop(); 33 logger.info("FetchConfirmOrdersFromErpJob end................."+ new Date()+" total cost time:"+watch.getTime()+" dealth data count:"+confirmOrderIds.size()); 34 }
OrderCalculate.java 队列获取订单线程
1 public class OrderCalculate { 2 3 private static final Log logger = LogFactory.getLog(OrderCalculate.class); 4 5 @Autowired 6 private static WriteRedisService writeRedisService; 7 8 private static ExecutorService threadPool=Executors.newFixedThreadPool(Runtime.getRuntime().availableProcessors()*4 9 ,new TjThreadFactory("CalculateAmount")); 10 static{ 11 Runtime.getRuntime().addShutdownHook(new Thread(new Runnable() { 12 @Override 13 public void run() { 14 QueuePop.stop(); 15 threadPool.shutdown(); 16 } 17 })); 18 } 19 20 public void init(){ 21 if(writeRedisService==null){ 22 writeRedisService=SpringContext.getBean(WriteRedisService.class); 23 } 24 new Thread(new QueuePop(),"OrderIdQueuePop").start();//由于是用redis做的队列，所以只要使用一个线程从队列里拿就ok 25 } 26 27 static class QueuePop implements Runnable{ 28 29 volatile static boolean stop=false; 30 31 @Override 32 public void run() { 33 while(!stop){ 34 //不断循环从队列里取出订单id 35 String orderId=null; 36 try { 37 orderId = writeRedisService.rpop(Constans.CACHE_ORDERIDS); 38 if(orderId!=null){ 39 logger.info("pop orderId:"+orderId);
//将获取的订单编号交给订单统计任务处理线程处理 40 threadPool.submit(new CalculateAmount(Integer.parseInt(orderId),new Date())); 41 } 42 } catch (Exception e1) { 43 logger.error("",e1); 44 } 45 //根据上线后的业务反馈来确定是否改成wait/notify策略来及时处理确认的订单 46 try { 47 Thread.sleep(10); 48 } catch (InterruptedException e) { 49 logger.error("",e); 50 // Thread.currentThread().interrupt(); 51 //stop=true;//线程被打算继续执行，不应该被关闭，保证该线程永远不会死掉 52 } 53 } 54 } 55 56 public static void stop(){ 57 stop=true; 58 } 59 60 } 61 62 }
CalculateAmoiunt.java 订单任务处理
1 public class CalculateAmount implements Runnable { 2 private static final Log logger = LogFactory.getLog(CalculateAmount.class); 3 private int orderId; 4 private Date now;//确认时间 这个时间有一定的延迟，基本可以忽略，如果没什么用 5 private OrderService orderServices; 6 private OrdHaveProductService ordHaveProductService; 7 private OrdPayByCashbackService ordPayByCashbackService; 8 private OrdPayByCouponService ordPayByCouponService; 9 private OrdPayByGiftCardService ordPayByGiftCardService; 10 private StatisticsService statisticsService; 11 private WriteRedisService writeRedisService; 12 private ReadRedisService readRedisService; 13 private ErpOrderGoodsService erpOrderGoodsService; 14 private ErpOrderService erpOrderService; 15 16 17 public CalculateAmount(int orderId,Date now) { 18 super(); 19 this.orderId = orderId; 20 this.now=now; 21 orderServices=SpringContext.getBean(OrderService.class); 22 ordHaveProductService=SpringContext.getBean(OrdHaveProductService.class); 23 ordPayByCashbackService=SpringContext.getBean(OrdPayByCashbackService.class); 24 ordPayByCouponService=SpringContext.getBean(OrdPayByCouponService.class); 25 ordPayByGiftCardService=SpringContext.getBean(OrdPayByGiftCardService.class); 26 statisticsService=SpringContext.getBean(StatisticsService.class); 27 writeRedisService=SpringContext.getBean(WriteRedisService.class); 28 readRedisService=SpringContext.getBean(ReadRedisService.class); 29 erpOrderGoodsService=SpringContext.getBean(ErpOrderGoodsService.class); 30 erpOrderService=SpringContext.getBean(ErpOrderService.class); 31 } 32 33 @Override 34 public void run() { 35 logger.info("CalculateAmount task run start........orderId:"+orderId); 36 StopWatch watch=new StopWatch(); 37 watch.start(); 38 /** 39 * 取出订单相关的所有数据同步到统计的库中 40 */ 41 //TODO 考虑要不要将下面所有操作放到一个事务里面 42 List<Map<String, Object>> orders = this.orderServices.selectOrderById(orderId); 43 if(orders!=null&&orders.size()>0){ 44 Map<String, Object> order = orders.get(0); 45 46 String orderSN=U.nvl(order.get("OrderSN"));//订单编号 47 Integer userId=U.nvlInt(order.get("usr_UserID"),null);//用户d 48 Integer status=U.nvlInt(order.get("Status"),null);//状态 49 Date createTime=now;//(Date)order.get("CreateTime");//创建时间 50 Date modifyTime=now;//(Date)order.get("ModifyTime");// 更新时间 51 BigDecimal discountPrice=U.nvlDecimal(order.get("DiscountPrice"),null);//优惠总额 满减金额 52 BigDecimal payPrice=U.nvlDecimal(order.get("PayPrice"), null);//实付金额 53 BigDecimal totalPrice=U.nvlDecimal(order.get("TotalPrice"), null);//总金额 54 55 //从erp里查询出订单的确认时间 56 int dbConfirmTime=0; 57 try { 58 dbConfirmTime = this.erpOrderService.selectConfirmTimeByOrderId(orderId); 59 } catch (Exception e2) { 60 logger.error("",e2); 61 } 62 Date ct=new Date(dbConfirmTime*1000L); 63 64 int[] dates=U.getYearMonthDayHour(ct);// 65 if(modifyTime!=null){ 66 dates=U.getYearMonthDayHour(modifyTime);// 67 } 68 int year=dates[0];//年 69 int month=dates[1];//月 70 int day=dates[2];//日 71 int hour=dates[3];//小时 72 73 String ordersId=orderId+"";//生成订单id 74 75 //查询订单的来源和搜索引擎关键字 76 String source=""; 77 String seKeyWords=""; 78 List<OrdersData> orderDataList=this.statisticsService.selectOrdersDataByOrdersId(orderSN); 79 if(orderDataList!=null&&!orderDataList.isEmpty()){ 80 OrdersData ordersData = orderDataList.get(0); 81 source=ordersData.getSource(); 82 seKeyWords=ordersData.getSeKeyWords(); 83 } 84 85 //TODO 将订单入库 86 ArrayList<RelOrders> relOrdersList = Lists.newArrayList(); 87 RelOrders relOrders=new RelOrders(orderSN,userId+"",Byte.valueOf(status+""),source,seKeyWords,IsCal.未计算.getFlag(),(byte)U.getSimpleYearByYear(year),(byte)month,(byte)day,(byte)hour,ct,createTime,modifyTime); 88 relOrdersList.add(relOrders); 89 90 try { 91 relOrders.setConfirmTime(ct); 92 //查询RelOrders是否存在 93 RelOrders dbOrders=this.statisticsService.selectByPrimaryKey(orderSN); 94 if(dbOrders!=null){ 95 //更新 96 dbOrders.setStatus(Byte.valueOf(status+"")); 97 dbOrders.setConfirmTime(ct); 98 dbOrders.setModifyTime(modifyTime); 99 this.statisticsService.updateByPrimaryKeySelective(dbOrders); 100 return; 101 }else{ 102 Integer relResult=this.statisticsService.insertRelOrdersBatch(relOrdersList); 103 } 104 } catch (Exception e) { 105 logger.error("insertRelOrdersBatch error",e); 106 } 107 /** 108 * 查这个订单的返现、优惠券、礼品卡 的金额 109 */ 110 List<Map<String, Object>> cashs = this.ordPayByCashbackService.selectDecutionPriceByOrderId(orderId); 111 List<Map<String, Object>> coupons = this.ordPayByCouponService.selectDecutionPriceByOrderId(orderId); 112 113 BigDecimal cashAmount=U.getValueByKey(cashs, "DeductionPrice", BigDecimal.class, BigDecimal.ZERO);//返现金额 114 BigDecimal couponAmont=U.getValueByKey(coupons, "DeductionPrice", BigDecimal.class, BigDecimal.ZERO);//红包金额 115 /** 116 * 查询出这个订单的所有商品 117 */ 118 List<Map<String, Object>> products=null; 119 Map<String,Object> productToKeyWordMap=Maps.newHashMap(); 120 try { 121 products = this.ordHaveProductService.selectByOrderId(orderId); 122 List<OrdersItemData> ordersItemDataList=this.statisticsService.selectOrdersItemDataByOrdersId(orderSN); 123 if(ordersItemDataList!=null){ 124 for (OrdersItemData ordersItemData : ordersItemDataList) { 125 productToKeyWordMap.put(ordersItemData.getItemId(), ordersItemData.getKeyWords()); 126 } 127 } 128 } catch (Exception e1) { 129 logger.error("",e1); 130 } 131 if(products!=null){ 132 ArrayList<RelOrdersItem> relOrdersItemList = Lists.newArrayList(); 133 for (Map<String, Object> product : products) { 134 Integer productId=U.nvlInt(product.get("pro_ProductID"), null);//商品Id 135 Integer buyNo=U.nvlInt(product.get("BuyNo"), 0);//购买数量 136 String SN=U.nvl(product.get("SN"),""); 137 BigDecimal buyPrice=U.nvlDecimal(product.get("BuyPrice"), BigDecimal.ZERO);//购买价格 138 BigDecimal buyTotalPrice=U.nvlDecimal(product.get("BuyTotalPrice"), null);//购买总价格 139 BigDecimal productPayPrice=U.nvlDecimal(product.get("PayPrice"), null);//单品实付金额 140 141 BigDecimal cost=null;//商品成本 TODO 调别人的接口 142 BigDecimal realtimeAmount=null;//实付金额 143 144 BigDecimal pdCashAmount=BigDecimal.ZERO;//每个商品的返现 145 BigDecimal pdcouponAmont=BigDecimal.ZERO;//每个商品的优惠券 146 147 //商品价格所占订单比例 148 if(buyTotalPrice!=null&&totalPrice!=null&&totalPrice.doubleValue()!=0){ 149 pdCashAmount=buyTotalPrice.divide(totalPrice,8,BigDecimal.ROUND_HALF_UP).multiply(cashAmount).setScale(2,BigDecimal.ROUND_HALF_UP); 150 pdcouponAmont=buyTotalPrice.divide(totalPrice,8,BigDecimal.ROUND_HALF_UP).multiply(couponAmont).setScale(2,BigDecimal.ROUND_HALF_UP); 151 discountPrice=buyTotalPrice.divide(totalPrice,8,BigDecimal.ROUND_HALF_UP).multiply(discountPrice).setScale(2,BigDecimal.ROUND_HALF_UP); 152 } 153 154 realtimeAmount=buyTotalPrice.subtract((pdCashAmount.add(pdcouponAmont).add(discountPrice))).setScale(2,BigDecimal.ROUND_HALF_UP); 155 156 RelOrdersItem item=new RelOrdersItem(U.randomUUID(),orderSN,productId,SN,buyNo,realtimeAmount,U.nvl(productToKeyWordMap.get(productId))); 157 158 relOrdersItemList.add(item); 159 160 //如果确认时间属于同一天的话，将商品实付金额放入到redis排行榜中 161 if((status==1||status==5||status==6||status==7||status==11)&&DateUtils.isSameDay(new Date(), ct)){ 162 //如果订单的状态是这几种，刚将该商品加入到实付金额的排行 榜中 163 dates=U.getYearMonthDayHour(ct);// 164 int days=dates[2]; 165 //某一个商品某一天的实付金额 166 BigDecimal itemRelAmount=BigDecimal.ZERO; 167 //从redis里取出这个商品的实付金额，然后累加 168 String itemRelAmountStr=readRedisService.get(Constans.CACHE_PERITEMRELAMOUNTSS_KEY_PREFIX+productId+Constans.CACHE_KEY_SEPARATOR+days); 169 if(StringUtils.isNotBlank(itemRelAmountStr)){ 170 itemRelAmount=new BigDecimal(itemRelAmountStr); 171 } 172 realtimeAmount=itemRelAmount.add(realtimeAmount); 173 writeRedisService.set(Constans.CACHE_PERITEMRELAMOUNTSS_KEY_PREFIX+productId+Constans.CACHE_KEY_SEPARATOR+days, realtimeAmount.toPlainString()); 174 writeRedisService.lpush(Constans.CACHE_DELKEYS_KEY_PRDFIX+days, Constans.CACHE_PERITEMRELAMOUNTSS_KEY_PREFIX+productId+Constans.CACHE_KEY_SEPARATOR+days); 175 writeRedisService.zadd(Constans.CACHE_ITEMREALAMOUNTSS_KEY+days, realtimeAmount.doubleValue(), productId+""); 176 //确认的销量 177 Long itemCount= writeRedisService.incrBy(Constans.CACHE_ITEMSALES_KEY_PRDFIX+productId+Constans.CACHE_KEY_SEPARATOR+days,buyNo); 178 writeRedisService.zadd(Constans.CACHE_ITEMSALES_SS_KEY_PRDFIX+days, itemCount, productId+""); 179 180 String itemType=""; 181 Map<String, String> pMap = this.readRedisService.hmget(Constans.CACHE_PRODUCT_KEY+productId); 182 itemType=pMap.get("categoryId"); 183 if(StringUtils.isNotBlank(itemType)){ 184 if(ProductCategory.isGuanBai(itemType)){ 185 //如果是白酒 官白的访客数排行 186 this.writeRedisService.zadd(Constans.CACHE_ITEMREALAMOUNTWHITESS_KEY+days, realtimeAmount.doubleValue(), productId+"");// 187 //确认的销量排行 188 this.writeRedisService.zadd(Constans.CACHE_ITEMSALESWHITE_SS_KEY_PRDFIX+days, itemCount, productId+"");// 189 }else if(ProductCategory.isGuanHong(itemType)){ 190 //官红的访客数排行 191 this.writeRedisService.zadd(Constans.CACHE_ITEMREALAMOUNTREDSS_KEY+days, realtimeAmount.doubleValue(), productId+"");// 192 //确认的销量排行 193 this.writeRedisService.zadd(Constans.CACHE_ITEMSALESRED_SS_KEY_PRDFIX+days, itemCount, productId+"");// 194 } 195 } 196 197 //某一个商品的销量加入删除列表 198 writeRedisService.lpush(Constans.CACHE_DELKEYS_KEY_PRDFIX+days, Constans.CACHE_ITEMSALES_KEY_PRDFIX+productId+Constans.CACHE_KEY_SEPARATOR+days); 199 } 200 } 201 try { 202 //TODO 将订单商品明细入库 203 this.statisticsService.insertRelOrdersItemBatch(relOrdersItemList); 204 //再将订单的状态改为已计算 205 this.statisticsService.updateIsCal(orderSN,IsCal.已计算.getFlag());//将是否计算改成已计算 206 //该订单的所有商品的成本同步到现在的库中。 207 this.calOrderProductCostSync(orderId,orderSN,products); 208 } catch (Exception e) { 209 logger.error("insertRelOrdersItemBatch or updateIsCal error",e); 210 } 211 } 212 } 213 watch.stop(); 214 logger.info("CalculateAmount task run end........total cost time:"+watch.getTime()+" orderId:"+orderId); 215 } 216 217 private void calOrderProductCostSync(int orderId,String orderSN,List<Map<String, Object>> products){ 218 List<Map<String, Object>> ordersList = this.erpOrderGoodsService.selectProductCostByOrderSN(orderSN); 219 if(ordersList==null||ordersList.isEmpty()){ 220 logger.error("according orderId to query some data from erp return is null........."); 221 return; 222 } 223 Map<String, String> itemIdToItemSnMap = U.convertToMapByList(products, "pro_ProductID", "SN"); 224 225 List<RelItemCosts> list=Lists.newArrayList(); 226 for (Map<String, Object> map : ordersList) { 227 RelItemCosts itemCost=new RelItemCosts(); 228 if(map==null){ 229 continue; 230 } 231 Integer itemId=U.nvlInt(map.get("goods_id"),-99); 232 BigDecimal costs=U.nvlDecimal(map.get("Dynamic_price"), BigDecimal.ZERO); 233 itemCost.setId(U.randomUUID()); 234 itemCost.setOrdersId(orderId+""); 235 itemCost.setOrdersNo(orderSN); 236 itemCost.setItemId(itemId); 237 itemCost.setItemNo(itemIdToItemSnMap.get(itemId+"")); 238 itemCost.setCosts(costs); 239 itemCost.setCreateTime(new Date()); 240 itemCost.setModifyTime(new Date()); 241 list.add(itemCost); 242 } 243 244 this.statisticsService.insertRelItemCostsBatch(list); 245 246 } 247 248 }
注意：
1、redis2.6版本使用lpush、rpop出列的时候会丢失数据。换成2.8及以上的版本运行正常。
2、由于应用会部署到多个结点，所以无法直接采用java的BlockingQueue阻塞队列，帮采用redis提供的队列支持。
3、如果要做到统计的绝对实时，最好采用大数据的实时计算的解决方案：kafka+storm 来实现
以上为队列结合线程的实践案例，供大家一起探讨。
转载请注明出处 ，请大家尊重作者的劳动成果。