Technology Sharing

Cache-Cache Usage 2

2024-07-12

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

1. Cache breakdown, penetration, and avalanche

1. Cache penetration

Refers to querying a data that does not exist. Since the cache does not hit, the database will be queried, but the database does not have such a record. We did not write the null of this query into the cache, which will cause this non-existent data to be queried at the storage layer every time it is requested, which defeats the purpose of caching.

risk:

Using non-existent data to attack, the database will be instantly stressed and eventually crash.

solve:

Null result cache, and add a short expiration time

2. Cache Avalanche

Cache avalanche: Cache avalanche means that when we set the cache, the keys use the same expiration time, causing the cache to become invalid at a certain moment, and all requests are forwarded to the DB, causing the DB to be overloaded and avalanche.

Solution: Add a random value to the original expiration time, such as 1-5 minutes. This way, the repetition rate of each cache expiration time will be reduced, making it difficult to cause collective failure events.

3. Cache breakdown

Cache breakdown:

  • For some keys with expiration times set, if these keys may be accessed with extremely high concurrency at certain points in time, they are very "hot" data.
  • If this hot key expires before a large number of requests come in at the same time, then all data queries for this key will fall into the DB, which we call cache breakdown.

solve:

Lock

In a large number of concurrent cases, only one person is allowed to check, and the others wait. After checking, the lock is released, and the others acquire the lock, hover, and check the cache first, and then there will be data without going to the DB.

2. Problem Solving

Solve cache penetration and avalanche first

  1.   private static final String CATALOG_JSON="CATALOG_JSON";
  2.     @Override
  3.     public Map<String, List<Catelog2Vo>> getCatalogJson() {
  4.  
  5.         /**
  6.          * 空结果缓存:解决缓存穿透
  7.          * 设置过期时间(加随机值) 缓存雪崩
  8.          * 加锁 解决缓存击穿
  9.          */
  10.         Object result = redisTemplate.opsForValue().get(CATALOG_JSON);
  11.         if(result!=null){
  12.             return (Map<String, List<Catelog2Vo>>) result;
  13.         }
  14.  
  15.         Map<String, List<Catelog2Vo>> map = getCatalogJsonFromDB();
  16.         if (map==null){
  17.             /**
  18.              * 解决缓存穿透
  19.              */
  20.             map=new HashMap<>();
  21.         }
  22.         redisTemplate.opsForValue().set(CATALOG_JSON,map, Duration.ofDays(1));
  23.         return map;
  24.     }

Solving cache breakdown

1. Use local lock solution

The springboot container object is in singleton mode by default, so you can synchronize the locks on the same object and execute concurrently when using the dual detection mode.

  1.  public synchronized Map<String, List<Catelog2Vo>> getCatalogJsonFromDB() {
  2.  
  3.         Object result = redisTemplate.opsForValue().get(CATALOG_JSON);
  4.         if (result != null) {
  5.             return (Map<String, List<Catelog2Vo>>) result;
  6.         }
  7.  
  8.  
  9.         //1.查出所有1级分类
  10.         List<CategoryEntity> selectList = baseMapper.selectList(null);
  11.         /**
  12.          * 将数据库的多次查询变成一次
  13.          */
  14.  
  15.         //2. 封装数据
  16.         List<CategoryEntity> level1Category = selectList.stream().filter(s -> s.getParentCid().equals(0L)).collect(Collectors.toList());
  17.         Map<String, List<Catelog2Vo>> map = level1Category.stream().collect(Collectors.toMap(k -> k.getCatId().toString(), v -> {
  18.             //1.每一个的一级分类,查到1级分类的所有二级分类
  19.             List<CategoryEntity> categoryEntities = selectList.stream().filter(s -> s.getParentCid().equals(v.getCatId())).collect(Collectors.toList());
  20.  
  21.             List<Catelog2Vo> catelog2VoList = categoryEntities.stream().map(c -> {
  22.                 Catelog2Vo catelog2Vo = new Catelog2Vo();
  23.                 catelog2Vo.setId(c.getCatId().toString());
  24.                 catelog2Vo.setName(c.getName());
  25.                 catelog2Vo.setCatalog1Id(v.getCatId().toString());
  26.  
  27.                 List<CategoryEntity> categoryEntities1 = selectList.stream().filter(s -> s.getParentCid().equals(c.getCatId())).collect(Collectors.toList());
  28.                 List<Catelog2Vo.Catelog3Vo> collect = categoryEntities1.stream().map(c3 -> {
  29.                     Catelog2Vo.Catelog3Vo catelog3Vo = new Catelog2Vo.Catelog3Vo();
  30.                     catelog3Vo.setId(c3.getCatId().toString());
  31.                     catelog3Vo.setName(c3.getName());
  32.                     catelog3Vo.setCatalog2Id(c.getCatId().toString());
  33.                     return catelog3Vo;
  34.                 }).collect(Collectors.toList());
  35.  
  36.                 catelog2Vo.setCatalog3List(collect);
  37.  
  38.                 return catelog2Vo;
  39.             }).collect(Collectors.toList());
  40.  
  41.  
  42.             return catelog2VoList;
  43.         }));
  44.         return map;
  45.     }
  1.  public Map<String, List<Catelog2Vo>> getCatalogJson() {
  2.  
  3.         /**
  4.          * 空结果缓存:解决缓存穿透
  5.          * 设置过期时间(加随机值) 缓存雪崩
  6.          * 加锁 解决缓存击穿
  7.          */
  8.         Object result = redisTemplate.opsForValue().get(CATALOG_JSON);
  9.         if (result != null) {
  10.             return (Map<String, List<Catelog2Vo>>) result;
  11.         }
  12.  
  13.         Map<String, List<Catelog2Vo>> map = getCatalogJsonFromDB();
  14.         if (map == null) {
  15.             /**
  16.              * 解决缓存穿透
  17.              */
  18.             map = new HashMap<>();
  19.         }
  20.         redisTemplate.opsForValue().set(CATALOG_JSON, map, Duration.ofDays(1));
  21.         return map;
  22.     }

The above code logic still has problems. When it is concurrent, thread 1 will release the lock before putting the data into the cache after checking the database. As a result, thread 2 will check the cache and find no data. It will check the database again. There is no guarantee that only one thread will check the database.

The right approach

  1.   public synchronized Map<String, List<Catelog2Vo>> getCatalogJsonFromDB() {
  2.  
  3.         Object result = redisTemplate.opsForValue().get(CATALOG_JSON);
  4.         if (result != null) {
  5.             return (Map<String, List<Catelog2Vo>>) result;
  6.         }
  7.  
  8.  
  9.         //1.查出所有1级分类
  10.         List<CategoryEntity> selectList = baseMapper.selectList(null);
  11.         /**
  12.          * 将数据库的多次查询变成一次
  13.          */
  14.  
  15.         //2. 封装数据
  16.         List<CategoryEntity> level1Category = selectList.stream().filter(s -> s.getParentCid().equals(0L)).collect(Collectors.toList());
  17.         Map<String, List<Catelog2Vo>> map = level1Category.stream().collect(Collectors.toMap(k -> k.getCatId().toString(), v -> {
  18.             //1.每一个的一级分类,查到1级分类的所有二级分类
  19.             List<CategoryEntity> categoryEntities = selectList.stream().filter(s -> s.getParentCid().equals(v.getCatId())).collect(Collectors.toList());
  20.  
  21.             List<Catelog2Vo> catelog2VoList = categoryEntities.stream().map(c -> {
  22.                 Catelog2Vo catelog2Vo = new Catelog2Vo();
  23.                 catelog2Vo.setId(c.getCatId().toString());
  24.                 catelog2Vo.setName(c.getName());
  25.                 catelog2Vo.setCatalog1Id(v.getCatId().toString());
  26.  
  27.                 List<CategoryEntity> categoryEntities1 = selectList.stream().filter(s -> s.getParentCid().equals(c.getCatId())).collect(Collectors.toList());
  28.                 List<Catelog2Vo.Catelog3Vo> collect = categoryEntities1.stream().map(c3 -> {
  29.                     Catelog2Vo.Catelog3Vo catelog3Vo = new Catelog2Vo.Catelog3Vo();
  30.                     catelog3Vo.setId(c3.getCatId().toString());
  31.                     catelog3Vo.setName(c3.getName());
  32.                     catelog3Vo.setCatalog2Id(c.getCatId().toString());
  33.                     return catelog3Vo;
  34.                 }).collect(Collectors.toList());
  35.  
  36.                 catelog2Vo.setCatalog3List(collect);
  37.  
  38.                 return catelog2Vo;
  39.             }).collect(Collectors.toList());
  40.  
  41.  
  42.             return catelog2VoList;
  43.         }));
  44.         if (map == null) {
  45.             /**
  46.              * 解决缓存穿透
  47.              */
  48.             map = new HashMap<>();
  49.         }
  50.         redisTemplate.opsForValue().set(CATALOG_JSON, map, Duration.ofDays(1));
  51.         return map;
  52.     }

Put the operation of storing in cache into the synchronized code block

  1.  @Override
  2.     public Map<String, List<Catelog2Vo>> getCatalogJson() {
  3.  
  4.         /**
  5.          * 空结果缓存:解决缓存穿透
  6.          * 设置过期时间(加随机值) 缓存雪崩
  7.          * 加锁 解决缓存击穿
  8.          */
  9.         Object result = redisTemplate.opsForValue().get(CATALOG_JSON);
  10.         if (result != null) {
  11.             return (Map<String, List<Catelog2Vo>>) result;
  12.         }
  13.  
  14.         Map<String, List<Catelog2Vo>> map = getCatalogJsonFromDB();
  15.  
  16.         return map;
  17.     }

 

 

Local locks can only lock the current process, so we need distributed locks

3. Problems with local locks in distributed environments

That is, each lock can only lock the current process, that is, each service will check the database once

Personal profile

I have devoted myself to the research of technology for more than 30 years. I am proficient in various languages ​​such as Java, Linux, JavaScript, PHP, CSS, etc. I have made many contributions in the field of open source. I have established a developer documentation site to share some problems in technology development for everyone to read.

my contact information

Mail