当前位置：网站首页>Multithreading uses redis to accumulate. The result is incorrect. Solution

Multithreading uses redis to accumulate. The result is incorrect. Solution

2022-04-22 16:36:00 【Time is an antidote】

multithreading redis The result of accumulation is not the right solution

The problem background
Solution
summary
Lyric： If you are sad, please forget me
This is the first 4 song , It's over , Did you guess the title of the song ？

The problem background

In the use of multithreading redis Of value An error occurred while accumulating , Use pseudo code for cause analysis

for(int i=0;i<10;i++){
    
	// Submit 10 Threads , Every params Equipped with 10 Data 
	asyncServiceExecutor.submit(() -> submitSingle(globalId, params));
}


public Map<String,String> submitSingle(String globalId, JSONArray params) {
    
    try {
    
        log.info("params: {}, size: {}", params, params.size());
        Map<String,String> resultVOlist = QueryController.doQuery(params);
        // Because it is multi-core , It can achieve real concurrency at the same time , Therefore, there may be 5 Threads arrive at this statement at the same time 
        // For the first time at the same time ,redis There is no such thing in it globalId keyword , therefore 5 All threads are for null
        Object commitObject = redisUtils.get(globalId);
        // When there is 5 A for null After the thread comes in , It will cause the same key to be created multiple times , Accumulating data is equivalent to losing 4 Time , That is to say 40 strip 
        if (commitObject == null) {
    
            boolean flag = redisUtils.set(globalId, params.size());
        } else {
    
            redisUtils.increment(globalId, (long) params.size());
        }
        return resultVOlist;
    } catch (Exception e) {
    
        log.error("QueryController not found error", e);
        return new HashMap<>();
    } 
}

Solution

1 Since it is the cause of multithreading , Then we can write redis The write operation of is moved out to the main thread ,redis The underlying write operation is single threaded

for(int i=0;i<10;i++){
    
		// Submit 10 Threads , Every params Equipped with 10 Data 
	Future<Map<String,String>> futureResult = asyncServiceExecutor.submit(() -> submitSingle(globalId, params));
	Map<String,String> resultVOList = futureResult.get();
	if(resultVOList != null){
    
		Object commitObject = redisUtils.get(globalId);
		if (commitObject == null) {
    
		    boolean flag = redisUtils.set(globalId, params.size());
		} else {
    
		    redisUtils.increment(globalId, (long) params.size());
		}
	}
}

public Map<String,String> submitSingle(String globalId, JSONArray params) {
    
    try {
    
        log.info("params: {}, size: {}", params, params.size());
        Map<String,String> resultVOlist = QueryController.doQuery(params);
        return resultVOlist;
    } catch (Exception e) {
    
        log.error("QueryController not found error", e);
        return new HashMap<>();
    } 
}

2 You can write... In the thread redis The logic of is locked as a whole

public Map<String,String> submitSingle(String globalId, JSONArray params) {
    
    try {
    
        log.info("params: {}, size: {}", params, params.size());
        Map<String,String> resultVOlist = QueryController.doQuery(params);
		if(redission.lock(uuid)){
    
	        Object commitObject = redisUtils.get(globalId);
	        if (commitObject == null) {
    
	            boolean flag = redisUtils.set(globalId, params.size());
	        } else {
    
	            redisUtils.increment(globalId, (long) params.size());
	        }
        }
        return resultVOlist;
    } catch (Exception e) {
    
        log.error("QueryController not found error", e);
        return new HashMap<>();
    } 
}