How to use Go and Lua to solve the inventory and overselling problems in Redis flash sale

0. Introduction

• Go language connects go-redis to connect to the database. If you don’t understand this part yet, it is recommended that you learn this part of the knowledge first.

• In addition, this flash sale mainly solves two problems. The first is the oversold problem, and the other is the inventory problem.

• There is no special page designed to simulate concurrency. We directly use gorountine and stay for 10 seconds before calling the request.

• To deal with the oversold problem, just introduce go-redis watch with transaction processing [equivalent to optimistic locking].

The inventory problem is a little more troublesome. You need to use Lua to edit the script, but you don't need to download the Lua compilation environment on your own machine. Go provides related support. For this part, don’t panic. The basic structure is as follows:

1. Simple version

In concurrent situations, oversold and negative values ​​may occur. Appear in the Redis database. Even if you make judgments on the data before operating.

func MsCode(uuid, prodid string) bool {  
   // 1、对uuid和prodid进行非空判断  
   if uuid == "" || prodid == "" {  
      return false  
   }  
  
   //2、获取连接  
   rdb := DB  
  
   //3、拼接key  
   kcKey := "kc:" + prodid + ":qt"  
   userKey := "sk:" + prodid + ":user"  
  
   //4、获取库存  
   str, err := rdb.Get(ctx, kcKey).Result()  
   if err != nil {  
      fmt.Println(err)  
      fmt.Println("秒杀还未开始.......")  
      return false  
   }  
  
   // 5、判断用户是否重复秒杀操作  
   flag, err := rdb.SIsMember(ctx, userKey, userKey).Result()  
   if err != nil {  
      fmt.Println(err)  
   }  
   if flag {  
      fmt.Println("你已经参加了秒杀，无法再次参加。。。。")  
      return false  
   }  
  
   // 6、判断商品数量，如果库存数量小于1，秒杀结束  
   str, err = rdb.Get(ctx, kcKey).Result()  
   if err != nil {  
      fmt.Println(err)  
   }  
   n, err := strconv.Atoi(str)  
   if err != nil {  
      fmt.Println(err)  
   }  
   if n < 1 {  
      fmt.Println("秒杀结束，请下次再来吧。。。。")  
      return false  
   }  
  
   // 7、秒杀过程  
   // 7.1、库存减1  
   num, err := rdb.Decr(ctx, kcKey).Result()  
   if err != nil {  
      fmt.Println(err)  
   }  
   if num != 0 {  
      // 7.2、添加用户  
      rdb.SAdd(ctx, userKey, uuid)  
   }  
   return true  
}

func main() {
    // 并发的版本
    for i := 0; i < 20; i++ {
        go func() {
            uuid := GenerateUUID()
            prodid := "1023"
            time.Sleep(10 * time.Second)
            MsCode(uuid, prodid)
        }()
    }
    time.Sleep(15 * time.Second)
}

2. Solve oversold

Use watch to monitor the key. The key parts are as follows. However, this situation will bring about a problem. Even if there is remaining inventory, there will be people who cannot buy it.

err = rdb.Watch(ctx, func(tx *redis.Tx) error {  
   n, err := tx.Get(ctx, kcKey).Int()  
   if err != nil && err != redis.Nil {  
      return err  
   }  
   if n <= 0 {  
      return fmt.Errorf("抢购结束了！请下次早点来。。。。")  
   }  
   _, err = tx.TxPipelined(ctx, func(pipeliner redis.Pipeliner) error {  
      err := pipeliner.Decr(ctx, kcKey).Err()  
      if err != nil {  
         return err  
      }  
      err = pipeliner.SAdd(ctx, userKey, uuid).Err()  
      if err != nil {  
         return err  
      }  
      return nil  
   })  
   return err  
}, kcKey)

3. Solve the inventory problem Lua

Lua can better solve this problem by operating redis. In order to avoid inventory problems that pessimistic locking may cause in Redis, you should consider using optimistic locking. Because Redis does not have built-in optimistic locking support, we need to use Lua to write relevant scripts. It mainly has the following advantages:

• Write complex or multi-step redis operations as a script and submit it to redis for execution at one time, reducing the number of repeated connections to redis. Improve performance.

• luan script is similar to a redis transaction, has a certain degree of atomicity, will not be queued by other commands, and can complete some redis transactional operations.

• The Lua script function of redis can only be used in versions above redis2.6.

• Use lua scripts to eliminate users and solve the overselling problem.

• After redis version 2.6, the contention problem is solved through Lua script. In fact, redis uses its single-threaded feature to solve multi-task concurrency problems using task queues.

import (
    "context"
    "fmt"
    "github.com/go-redis/redis/v8"
    "net"
    "time"
)

func useLua(userid, prodid string) bool {
    //编写脚本 - 检查数值，是否够用，够用再减，否则返回减掉后的结果
    var luaScript = redis.NewScript(`
        local userid=KEYS[1];
        local prodid=KEYS[2];
        local qtKey="sk:"..prodid..":qt";
        local userKey="sk:"..prodid..":user";
        local userExists=redis.call("sismember",userKey,userid);
        if tonumber(userExists)==1 then
         return 2;
        end
        local num=redis.call("get",qtKey);
        if tonumber(num)<=0 then
         return 0;
        else
         redis.call("decr",qtKey);
         redis.call("SAdd",userKey,userid);
        end
        return 1;
    `)
    //执行脚本
    n, err := luaScript.Run(ctx, DB, []string{userid, prodid}).Result()
    if err != nil {
        return false
    }
    switch n {
    case int64(0):
        fmt.Println("抢购结束")
        return false
    case int64(1):
        fmt.Println(userid, "：抢购成功")
        return true
    case int64(2):
        fmt.Println(userid, "：已经抢购了")
        return false
    default:
        fmt.Println("发生未知错误！")
        return false
    }
    return true
}

func main() {
    // 并发的版本
    for i := 0; i < 20; i++ {
        go func() {
            uuid := GenerateUUID()
            prodid := "1023"
            time.Sleep(10 * time.Second)
            useLua(uuid, prodid)
        }()
    }
    time.Sleep(15 * time.Second)
}

The above is the detailed content of How to use Go and Lua to solve the inventory and overselling problems in Redis flash sale. For more information, please follow other related articles on the PHP Chinese website!