-
Record a Redis divine operation - also known as pit stepping diary The preface has not updated the blog for a long time, and I have updated it again. The problem encountered this time is to delete keys and query keys in Redis. The project background is the old rule. Let's start with the project background. The original mall project was opened for the second time, and the customer requested to add a nearby function. We already have the nearby merchant function, so we just need to open it to go up. If the question is so simple, then I don't have any articles to write? Of course not. After the project is put into the test environment, the product is tested, and the feedback is that the banner and commodity list on the home page cannot be managed. My first reaction is that this function is good. It must be that the new product doesn't understand anything. But in order to prevent being slapped on the face, I made a special attempt, and the result was really problematic. View the code. It is found that Redis has been deleted in the background, but the deletion is not complete. Because N programmers have developed this block, the function of this block is extremely crowded, like shit, and various operations. Now the demand is that each city and each region has a corresponding merchant to choose from. If there is no merchant, a designated city will be selected by default. A programmer wrote a magic operation: if there is no data of a certain city in the background, take the data of the default city and put it in the city, then a magic thing happened here: when the background manages the city data, such as updating or deleting the city data, but there is no data of the city in the front end, This causes the background to delete the relevant Redis, but not the existing Redis. For example, the user requests the data of Beijing in the front end, but there is no Beijing in the back end, so the data of Shenzhen is given to Beijing by default, and is cached in the data whose key is Beijing in Redis. At this time, when updating Shenzhen or adding Beijing in the back end, Redis does not do the corresponding update operation. Solution How to solve this problem? For the sake of speed, I added a set directly. Set a set, and put the nonexistent keys in the set, so that each time you can judge whether they are not in the set, and when you delete them in the background, you can judge whether the corresponding city keys exist in the set, and delete the corresponding values in the set if they exist. Other logic shall be executed as usual. For example, the existing city cache is directly deleted (unlink). Why don't you write in the background? Because this key may not have access to the afterword efficiency and function analysis for a long time, when changing this logic, various considerations were taken, and this method was finally used. The reason is that the set is a little less repetitive than the list, that is, the data duplication problem is ignored. Why not use an ordered set? Because you don't need to sort, you just need to know whether it exists. Business analysis: the original version: the corresponding city Redis cache cannot be deleted. The first version is modified: the collection is added. When deleting in the background, use hscan+unlink to delete the collection and the cached city key, and then the original logic remains unchanged. Fastest implementation. At this time, the front end requests that the city data that does not exist in the background will be cached in memory, causing a waste of memory. The second version has been modified: the front end first judges whether the city data exists. If not, it judges whether the city value exists in the collection. If it exists, it returns the default city cache. The background only needs to delete the new values in the collection. Thank you for your article reference https://blog.csdn.net/u012998306/article/details/105417611 Redis sadd command and delete the set set or delete one or more records in the set separately https://bbs.huaweicloud.com/blogs/317523 In order to grasp the Redis data structure, I drew 40 pictures (full version) https://blog.csdn.net/fuzhongmin05/article/details/119531696 How Data Structure Makes Redis Fast http://jinguoxing.github.io/redis/2018/09/04/redis-scan/ Use of scan https://www.lixueduan.com/post/redis/redis-scan/ Redis Scan principle analysis and pit stepping https://nyogjtrc.github.io/posts/2019/12/delete-keys-matching-a-pattern-in-redis/ Delete K conforming to Pattern in Redis eyhttps://redis.io/commands/unlink/ unlink http://mysql.taobao.org/monthly/2018/10/05/ Redis lazyfree · Gospel deleted by big key https://blog.csdn.net/loophome/article/details/105204427 Redis Delete Large Collection Operation (Lists, Sets) -
