Redistemplate solves the problem of oversold inventory in the seckill system with high speed - redis transaction + optimistic lock mechanism

1、 scene

Seckill system has high concurrency scenario , In the second kill of goods , High concurrency may lead to oversold inventory , So you can go through Redis The transaction mechanism provided is oversold ;Redis A transaction is actually executing all commands in sequence . The transaction will not be interrupted by other commands .

2、 Repeat the oversold scene

2.1 Initialize the inventory interface

@RestController
@RequestMapping("/redis")
@Slf4j
public class RedisController {
    
	@Resource
    private RedisTemplate redisTemplate;
	
	// Record the quantity of goods actually sold 
    private AtomicInteger successNum = new AtomicInteger(0);
    
	@GetMapping(value = "/init")
    public String init() {
    
        //  Initialize inventory quantity , Simulated inventory as long as 5 A commodity , Write to redis in 
        redisTemplate.opsForValue().set("stock", 5);
		successNum.set(0);
		
        log.info("===>>> Inventory initialization succeeded , The stock is " + 5);
        return " Inventory initialization succeeded ";
    }
}

2.2 Inventory deduction interface

@RestController
@RequestMapping("/redis")
@Slf4j
public class RedisController {
    
	@Resource
    private RedisTemplate redisTemplate;
    
    // Record the quantity of goods actually sold 
    private AtomicInteger successNum = new AtomicInteger(0);
    
	@GetMapping(value = "/reduce")
    public String reduce() {
    
        int stock = (Integer) redisTemplate.opsForValue().get("stock");
        log.info("===>>> Current quantity " + stock);
        //  The simulation reduces only one inventory 
        stock = stock - 1;
        if (stock < 0) {
    
            log.info("===>>> Insufficient inventory ");
            return " Insufficient inventory ";
        }
        //  Write the remaining quantity back to redis
        redisTemplate.opsForValue().set("stock", stock);
        //  Record the quantity of goods actually sold （ Thread safe every request will be logged ）
        log.info("===>>> Reduce inventory successfully , Sell... Together " + successNum.incrementAndGet());
        return " Reduce inventory successfully ";
    }
}

2.3 test

Using tools JMeter Simulate concurrent requests , Simulate here Per second 200 Time ;JMeter Tool use reference blog ：https://blog.csdn.net/tianqingmuyu/article/details/108401543

Be careful ： Perform... Before testing Initialize the inventory interface , Ensure inventory is written to Redis in
Use JMeter Request interface , The results are as follows ：

3、 Solve oversold and realize

3.1 Initialize the inventory interface

@RestController
@RequestMapping("/redis")
@Slf4j
public class RedisController {
    
	@Resource
    private RedisTemplate redisTemplate;
	
	// Record the quantity of goods actually sold 
    private AtomicInteger successNum = new AtomicInteger(0);
    
	@GetMapping(value = "/init")
    public String init() {
    
        //  Initialize inventory quantity , Simulated inventory as long as 5 A commodity , Write to redis in 
        redisTemplate.opsForValue().set("stock", 5);
		successNum.set(0);
		
        log.info("===>>> Inventory initialization succeeded , The stock is " + 5);
        return " Inventory initialization succeeded ";
    }
}

3.2 Inventory deduction interface

@RestController
@RequestMapping("/redis")
@Slf4j
public class RedisController {
    
	@Resource
    private RedisTemplate redisTemplate;
    
	// Record the quantity of goods actually sold 
    private AtomicInteger successNum = new AtomicInteger(0);
    
	@GetMapping(value = "/reduce")
    public String reduce() {
    
        //  Open transaction 
        redisTemplate.setEnableTransactionSupport(true);
        List<Object> results = (List<Object>) redisTemplate.execute(new SessionCallback<List<Object>>() {
    
            @Override
            public List<Object> execute(RedisOperations operations) throws DataAccessException {
    
                //  monitor key
                operations.watch("stock");
                Integer stock = (Integer) operations.opsForValue().get("stock");
                operations.multi();
                stock = stock - 1;
                if (stock < 0) {
    
                    log.info("===>>> Insufficient inventory ");
                    return null;
                }
                operations.opsForValue().set("stock", stock);
                return operations.exec();
            }
        });
        if (results != null && results.size() > 0) {
    
            log.info("===>>> Reduce inventory successfully , Sell... Together " + successNum.incrementAndGet());
            return " Reduce inventory successfully ";
        }
        return " Insufficient inventory ";
    }
}

3.3 test

Using tools JMeter Simulate concurrent requests , Simulate here Per second 200 Time ;JMeter Tool use reference blog ：https://blog.csdn.net/tianqingmuyu/article/details/108401543

Be careful ： Perform... Before testing Initialize the inventory interface , Ensure inventory is written to Redis in
Use JMeter Request interface , The results are as follows , There is no oversold ：

Conclusion
adopt Redis Transaction mechanism , It can effectively solve the oversold problem of the second kill system ;

Other implementations ：《RedisTemplate Solve the problem of oversold inventory in the second kill system — Redis Implement distributed locking mechanism 》

Statement ：JMeter Tool use reference blog ：https://blog.csdn.net/tianqingmuyu/article/details/108401543 — @ front end SkyRain