As mentioned earlier mysql Caching scheme for , Put the reading pressure on redis

Under normal circumstances , There is absolutely no problem with the cooperation between the two

Cache penetration

Suppose a data redis non-existent , mysql There is no such thing as ,⽽ And ⼀ Try to read straight ？ Cache penetration , The data is finally compressed ⼒
Still piled up in mysql, May cause mysql Overwhelmed ⽽ collapse ;

solve

1. Find out mysql non-existent , take redis Set to <key, nil> Set expiration time Next visit key When No longer visit
mysql Easy to cause redis Cache a lot ⽆ Effective data ;
2. The bloon filter , take mysql Which already exists key, Write ⼊ The bloon filter , Nonexistent direct pass fall ;

Cache breakdown

Cache breakdown Some data redis No, , however mysql Yes ; At the moment when ⼤ Concurrent requests for such data , It also causes mysql
too ⼤;

solve

1. Lock
Get the lock when requesting data , If successful , Operation , Acquisition failure , Then sleep ⼀ Period of time （200ms） Get it again
take ; To be successful , Then release the lock
⾸ First reading redis, non-existent , read mysql, There is , Write redis key Lock of
The whole process ⾛ End , Just let me ⾯ Server access
2. It will be very hot key, Set not to expire

Cache avalanche

Express ⼀ For a while , Cache set failure (redis⽆ mysql Yes ), Cause all requests to ⾛mysql, It's possible to bring down the database ,
Disable the entire service ;

solve

1. If the cache database goes down , Cause all data to flow to mysql;
Mining ⽤⾼ can ⽤ The cluster of ⽅ case , Like sentinel mode 、 cluster Pattern ;
2. If you set the same expiration time , Cause cache set invalidation ;
Set random expiration value or other mechanism to stagger failure ;
3. If the system restarts , Cause the cached data to disappear ;
Short restart time , redis Turn on persistence （ Expired information is also persisted ） Just ⾏ 了 ; Restart time ⻓ Guide the thermal data in advance
⼊redis among ;