Redis 缓存的三大问题及其解决方案

{"type":"doc","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Redis经常用于系统中的缓存，这样可以解决目前IO设备无法满足互联网应用海量的读写请求的问题。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"一、缓存穿透","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"缓存穿透是指缓存和数据库中都没有的数据，而用户不断发起请求，如发起id为-1的数据或者特别大的不存在的数据。有可能是黑客利用漏洞攻击从而去压垮应用的数据库。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"1. 常见解决方案","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"对于缓存穿透问题，常见的解决方案有以下三种：","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"验证拦截：接口层进行校验，如鉴定用户权限，对ID之类的字段做基础的校验，如id<=0的字段直接拦截；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"缓存空数据：当数据库查询到的数据为空时，也将这条数据进行缓存，但缓存的有效性设置得要较短，以免影响正常数据的缓存；","attrs":{}}]}]}],"attrs":{}},{"type":"codeblock","attrs":{"lang":"java"},"content":[{"type":"text","text":"Copypublic Student getStudentsByID(Long id) {\n\n // 从Redis中获取学生信息\n Student student = redisTemplate.opsForValue()\n .get(String.valueOf(id));\n if (student != null) {\n return student;\n }\n\n // 从数据库查询学生信息，并存入Redis\n student = studentDao.selectByStudentId(id);\n if (student != null) {\n redisTemplate.opsForValue()\n .set(String.valueOf(id), student, 60, TimeUnit.MINUTES);\n } else {\n // 即使不存在，也将其存入缓存中\n redisTemplate.opsForValue()\n .set(String.valueOf(id), null, 60, TimeUnit.SECONDS);\n }\n\n return student;\n}","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"使用布隆过滤器：布隆过滤器是一种比较独特数据结构，有一定的误差。当它指定一个数据存在时，它不一定存在，但是当它指定一个数据不存在时，那么它一定是不存在的。","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"2. 布隆过滤器","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"布隆过滤器是一种比较特殊的数据结构，有点类似与HashMap，在业务中我们可能会通过使用HashMap来判断一个值是否存在，它可以在O(1)时间复杂度内返回结果，效率极高，但是受限于存储容量，如果可能需要去判断的值超过亿级别，那么HashMap所占的内存就很可观了。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而BloomFilter解决这个问题的方案很简单。首先用多个bit位去代替HashMap中的数组，这样的话储存空间就下来了，之后就是对 Key 进行多次哈希，将 Key 哈希后的值所对应的 bit 位置为1。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"当判断一个元素是否存在时，就去判断这个值哈希出来的比特位是否都为1，如果都为1，那么可能存在，也可能不存在（如下图F）。但是如果有一个bit位不为1，那么这个Key就肯定不存在。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/85/851a1cbd67e8f4d1bb43179b363982dd.webp","alt":"图片","title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"注意：BloomFilter并不支持删除操作，只支持添加操作。这一点很容易理解，因为你如果要删除数据，就得将对应的bit位置为0，但是你这个Key对应的bit位可能其他的Key也对应着。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"3. 缓存空数据与布隆过滤器的比较","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"上面对这两种方案都进行了简单的介绍，缓存空数据与布隆过滤器都能有效解决缓存穿透问题，但使用场景有着些许不同；","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"当一些恶意攻击查询查询的key各不相同，而且数量巨多，此时缓存空数据不是一个好的解决方案。因为它需要存储所有的Key，内存空间占用高。并且在这种情况下，很多key可能只用一次，所以存储下来没有意义。所以对于这种情况而言，使用布隆过滤器是个不错的选择；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而对与空数据的Key数量有限、Key重复请求效率较高的场景而言，可以选择缓存空数据的方案。","attrs":{}}]}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"二、缓存击穿缓","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"存击穿是指当前热点数据存储到期时，多个线程同时并发访问热点数据。因为缓存刚过期，所有并发请求都会到数据库中查询数据。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"解决方案","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"将热点数据设置为永不过期；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"加互斥锁：互斥锁可以控制查询数据库的线程访问，但这种方案会导致系统的吞吐量下降，需要根据实际情况使用。","attrs":{}}]}]}],"attrs":{}},{"type":"codeblock","attrs":{"lang":"java"},"content":[{"type":"text","text":"Copypublic String get(key) {\n String value = redis.get(key);\n if (value == null) { // 代表缓存值过期\n // 设置3min的超时，防止del操作失败的时候，下次缓存过期一直不能load db\n if (redis.setnx(key_mutex, 1, 3 * 60) == 1) { // 代表设置成功\n value = db.get(key);\n redis.set(key, value, expire_secs);\n redis.del(key_mutex);\n } else { // 这个时候代表同时候的其他线程已经load db并回设到缓存了，这时候重试获取缓存值即可\n sleep(50);\n get(key); // 重试\n }\n } else {\n return value;\n }\n}\n","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"三、缓存雪崩","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"缓存雪崩发生有几种情况，比如大量缓存集中在或者缓存同时在大范围中失效，出现了大量请求去访问数据库，从而导致CPU和内存过载，甚至停机。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"一个简单的雪崩过程：","attrs":{}}]},{"type":"numberedlist","attrs":{"start":null,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","text":"Redis 集群产生了大面积故障；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","text":"缓存失败，此时仍有大量请求去访问 Redis 缓存服务器；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","text":"在大量 Redis 请求失败后，这些请求将会去访问数据库；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":4,"align":null,"origin":null},"content":[{"type":"text","text":"由于应用的设计依赖于数据库和 Redis 服务，很快就会造成服务器集群的雪崩，最终导致整个系统的瘫痪。","attrs":{}}]}]}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"解决方案","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"【事前】高可用缓存：高可用缓存是防止出现整个缓存故障。即使个别节点，机器甚至机房都关闭，系统仍然可以提供服务，Redis 哨兵(Sentinel) 和 Redis 集群(Cluster) 都可以做到高可用；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"【事中】缓存降级（临时支持）：当访问次数急剧增加导致服务出现问题时，我们如何确保服务仍然可用。在国内使用比较多的是 Hystrix，它通过熔断、降级、限流三个手段来降低雪崩发生后的损失。只要确保数据库不死，系统总可以响应请求，每年的春节 12306 我们不都是这么过来的吗？只要还可以响应起码还有抢到票的机会；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"【事后】Redis备份和快速预热：Redis数据备份和恢复、快速缓存预热。","attrs":{}}]}]}],"attrs":{}}]}