Redis+Caffeine 实现两级缓存实战

Redis+Caffeine 实现两级缓存

背景

​ 事情的开始是这样的，前段时间接了个需求，给公司的商城官网提供一个查询预计送达时间的接口。接口很简单，根据请求传的城市+仓库+发货时间查询快递的预计送达时间。因为商城下单就会调用这个接口，所以对接口的性能要求还是挺高的，据老员工的说法是特别是大促的时候，访问量还是比较大的。

​ 因为数据量不是很大，每天会全量推今天和明天的预计送达时间到MySQL，总数据量大约7k+。每次推完数据后会把数据全量写入到redis中，做一个缓存预热，然后设置过期时间为1天。

​ 鉴于之前Redis集群出现过压力过大查询缓慢的情况，进一步保证接口的高性能和高可用，防止redis出现压力大，查询慢，缓存雪崩，缓存穿透等问题，我们最终采用了Reids + Caffeine两级缓存的策略。

本地缓存优缺点

优点：

1. 本地缓存，基于本地内存，查询速度是很快的。适用于：实时性要求不高，更新频率不高等场景。（我们的数据每天凌晨更新一次，总量7k左右）
2. 查询本地缓存与查询远程缓存相比可以减少网络的I/O，降低网络上的一些消耗。（我们的redis之前出现过查询缓慢的情况）

缺点：

1. Caffeine既然是本地缓存，在分布式环境的情况下就要考虑各个节点之间缓存的一致性问题，一个节点的本地缓存更新了，怎么可以同步到其他的节点。
2. Caffeine不支持持久化的存储。
3. Caffeine使用本地内存，需要合理设置大小，避免内存溢出。

流程图

代码实现

MySQL表

CREATE TABLE `t_estimated_arrival_date`  (`id` int(11) UNSIGNED NOT NULL AUTO_INCREMENT COMMENT '主键id',`warehouse_id` varchar(32) CHARACTER SET utf8mb4 COLLATE utf8mb4_general_ci NULL DEFAULT NULL COMMENT '货仓id',`warehouse` varchar(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_general_ci NULL DEFAULT NULL COMMENT '发货仓',`city` varchar(255) CHARACTER SET utf8mb4 COLLATE utf8mb4_general_ci NULL DEFAULT NULL COMMENT '签收城市',`delivery_date` date NULL DEFAULT NULL COMMENT '发货时间',`estimated_arrival_date` date NULL DEFAULT NULL COMMENT '预计到货日期',PRIMARY KEY (`id`) USING BTREE,UNIQUE INDEX `uk_warehouse_id_city_delivery_date`(`warehouse_id`, `city`, `delivery_date`) USING BTREE
) ENGINE = InnoDB  COMMENT = '预计到货时间表(具体到day：T, T+1,近90天到货时间众数)' ROW_FORMAT = Dynamic;INSERT INTO `t_estimated_arrival_date` VALUES (9, '6', '湖熟正常仓', '兰州市', '2024-07-08', '2024-07-10');
INSERT INTO `t_estimated_arrival_date` VALUES (10, '6', '湖熟正常仓', '兰州市', '2024-07-09', '2024-07-11');
INSERT INTO `t_estimated_arrival_date` VALUES (11, '6', '湖熟正常仓', '兴安盟', '2024-07-08', '2024-07-11');
INSERT INTO `t_estimated_arrival_date` VALUES (12, '6', '湖熟正常仓', '兴安盟', '2024-07-09', '2024-07-12');
INSERT INTO `t_estimated_arrival_date` VALUES (13, '6', '湖熟正常仓', '其他', '2024-07-08', '2024-07-19');
INSERT INTO `t_estimated_arrival_date` VALUES (14, '6', '湖熟正常仓', '其他', '2024-07-09', '2024-07-20');
INSERT INTO `t_estimated_arrival_date` VALUES (15, '6', '湖熟正常仓', '内江市', '2024-07-08', '2024-07-10');
INSERT INTO `t_estimated_arrival_date` VALUES (16, '6', '湖熟正常仓', '内江市', '2024-07-09', '2024-07-11');
INSERT INTO `t_estimated_arrival_date` VALUES (17, '6', '湖熟正常仓', '凉山彝族自治州', '2024-07-08', '2024-07-11');
INSERT INTO `t_estimated_arrival_date` VALUES (18, '6', '湖熟正常仓', '凉山彝族自治州', '2024-07-09', '2024-07-12');
INSERT INTO `t_estimated_arrival_date` VALUES (19, '6', '湖熟正常仓', '包头市', '2024-07-08', '2024-07-11');
INSERT INTO `t_estimated_arrival_date` VALUES (20, '6', '湖熟正常仓', '包头市', '2024-07-09', '2024-07-12');
INSERT INTO `t_estimated_arrival_date` VALUES (21, '6', '湖熟正常仓', '北京城区', '2024-07-08', '2024-07-10');
INSERT INTO `t_estimated_arrival_date` VALUES (22, '6', '湖熟正常仓', '北京城区', '2024-07-09', '2024-07-11');

pom.xm

        <dependency><groupId>org.springframework.boot</groupId><artifactId>spring-boot-starter-aop</artifactId></dependency><dependency><groupId>org.springframework.boot</groupId><artifactId>spring-boot-starter-data-redis</artifactId></dependency><!--redis连接池--><dependency><groupId>org.apache.commons</groupId><artifactId>commons-pool2</artifactId></dependency><dependency><groupId>org.springframework.boot</groupId><artifactId>spring-boot-starter-cache</artifactId></dependency><dependency><groupId>com.github.ben-manes.caffeine</groupId><artifactId>caffeine</artifactId><version>2.9.2</version></dependency><dependency><groupId>mysql</groupId><artifactId>mysql-connector-java</artifactId><version>8.0.28</version></dependency><dependency><groupId>com.baomidou</groupId><artifactId>mybatis-plus-boot-starter</artifactId><version>3.3.1</version></dependency>

application.yml

server:port: 9001
spring:application:name: springboot-redisdatasource:name: demourl: jdbc:mysql://localhost:3306/test?userUnicode=true&&characterEncoding=utf8&allowMultiQueries=true&useSSL=falsedriver-class-name: com.mysql.cj.jdbc.Driverusername: password: # mybatis相关配置mybatis-plus:mapper-locations: classpath:mapper/*.xmlconfiguration:cache-enabled: trueuse-generated-keys: truedefault-executor-type: REUSEuse-actual-param-name: true# 打印日志#    log-impl: org.apache.ibatis.logging.stdout.StdOutImplredis:host: 192.168.117.73port: 6379password: root
#  redis:
#    lettuce:
#      cluster:
#        refresh:
#          adaptive: true
#          period: 10S
#      pool:
#        max-idle: 50
#        min-idle: 8
#        max-active: 100
#        max-wait: -1
#    timeout: 100000
#    cluster:
#      nodes:
#        - 192.168.117.73:6379
logging:level:com.itender.redis.mapper: debug

配置类

• RedisConfig

/*** @author yuanhewei* @date 2024/5/31 16:18* @description*/
@Configuration
public class RedisConfig {@Beanpublic RedisTemplate<String, Object> redisTemplate(RedisConnectionFactory connectionFactory) {RedisTemplate<String, Object> redisTemplate = new RedisTemplate<>();redisTemplate.setConnectionFactory(connectionFactory);Jackson2JsonRedisSerializer<Object> serializer = new Jackson2JsonRedisSerializer<>(Object.class);ObjectMapper mapper = new ObjectMapper();mapper.setVisibility(PropertyAccessor.ALL, JsonAutoDetect.Visibility.ANY);mapper.activateDefaultTyping(LaissezFaireSubTypeValidator.instance, ObjectMapper.DefaultTyping.NON_FINAL, JsonTypeInfo.As.PROPERTY);serializer.setObjectMapper(mapper);// 如果不序列化在key value 使用redis客户端工具 直连redis服务器 查看数据时 前面会有一个 \xac\xed\x00\x05t\x00\x05 字符串// StringRedisSerializer 来序列化和反序列化 String 类型 redis 的 key valueredisTemplate.setKeySerializer(new StringRedisSerializer());redisTemplate.setValueSerializer(serializer);// StringRedisSerializer 来序列化和反序列化 hash 类型 redis 的 key valueredisTemplate.setHashKeySerializer(new StringRedisSerializer());redisTemplate.setHashValueSerializer(serializer);redisTemplate.afterPropertiesSet();return redisTemplate;}
}

• CaffeineConfig

/*** @author yuanhewei* @date 2024/7/9 14:16* @description*/
@Configuration
public class CaffeineConfig {/*** Caffeine 配置类*  initialCapacity：初始缓存空间大小*  maximumSize：缓存的最大数量，设置这个值避免内存溢出*  expireAfterWrite：指定缓存的过期时间，是最后一次写操作的一个时间*  容量的大小要根据自己的实际应用场景设置** @return*/@Beanpublic Cache<String, Object> caffeineCache() {return Caffeine.newBuilder()// 初始大小.initialCapacity(128)//最大数量.maximumSize(1024)//过期时间.expireAfterWrite(60, TimeUnit.SECONDS).build();}@Beanpublic CacheManager cacheManager(){CaffeineCacheManager cacheManager=new CaffeineCacheManager();cacheManager.setCaffeine(Caffeine.newBuilder().initialCapacity(128).maximumSize(1024).expireAfterWrite(60, TimeUnit.SECONDS));return cacheManager;}
}

Mapper

这里采用了Mybatis Plus

/*** @author yuanhewei* @date 2024/7/9 18:11* @description*/
@Mapper
public interface EstimatedArrivalDateMapper extends BaseMapper<EstimatedArrivalDateEntity> {} 

Service

/*** @author yuanhewei* @date 2024/7/9 14:25* @description*/
public interface DoubleCacheService {/*** 查询一级送达时间-常规方式** @param request* @return*/EstimatedArrivalDateEntity getEstimatedArrivalDateCommon(EstimatedArrivalDateEntity request);/*** 查询一级送达时间-注解方式** @param request* @return*/EstimatedArrivalDateEntity getEstimatedArrivalDate(EstimatedArrivalDateEntity request);
}

实现类

/*** @author yuanhewei* @date 2024/7/9 14:26* @description*/
@Slf4j
@Service
public class DoubleCacheServiceImpl implements DoubleCacheService {@Resourceprivate Cache<String, Object> caffeineCache;@Resourceprivate RedisTemplate<String, Object> redisTemplate;@Resourceprivate EstimatedArrivalDateMapper estimatedArrivalDateMapper;@Overridepublic EstimatedArrivalDateEntity getEstimatedArrivalDateCommon(EstimatedArrivalDateEntity request) {String key = request.getDeliveryDate() + RedisConstants.COLON + request.getWarehouseId() + RedisConstants.COLON + request.getCity();log.info("Cache key: {}", key);Object value = caffeineCache.getIfPresent(key);if (Objects.nonNull(value)) {log.info("get from caffeine");return EstimatedArrivalDateEntity.builder().estimatedArrivalDate(value.toString()).build();}value = redisTemplate.opsForValue().get(key);if (Objects.nonNull(value)) {log.info("get from redis");caffeineCache.put(key, value);return EstimatedArrivalDateEntity.builder().estimatedArrivalDate(value.toString()).build();}log.info("get from mysql");DateTime deliveryDate = DateUtil.parse(request.getDeliveryDate(), "yyyy-MM-dd");EstimatedArrivalDateEntity estimatedArrivalDateEntity = estimatedArrivalDateMapper.selectOne(new QueryWrapper<EstimatedArrivalDateEntity>().eq("delivery_date", deliveryDate).eq("warehouse_id", request.getWarehouseId()).eq("city", request.getCity()));redisTemplate.opsForValue().set(key, estimatedArrivalDateEntity.getEstimatedArrivalDate(), 120, TimeUnit.SECONDS);caffeineCache.put(key, estimatedArrivalDateEntity.getEstimatedArrivalDate());return EstimatedArrivalDateEntity.builder().estimatedArrivalDate(estimatedArrivalDateEntity.getEstimatedArrivalDate()).build();}@DoubleCache(cacheName = "estimatedArrivalDate", key = {"#request.deliveryDate", "#request.warehouseId", "#request.city"},type = DoubleCache.CacheType.FULL)@Overridepublic EstimatedArrivalDateEntity getEstimatedArrivalDate(EstimatedArrivalDateEntity request) {DateTime deliveryDate = DateUtil.parse(request.getDeliveryDate(), "yyyy-MM-dd");EstimatedArrivalDateEntity estimatedArrivalDateEntity = estimatedArrivalDateMapper.selectOne(new QueryWrapper<EstimatedArrivalDateEntity>().eq("delivery_date", deliveryDate).eq("warehouse_id", request.getWarehouseId()).eq("city", request.getCity()));return EstimatedArrivalDateEntity.builder().estimatedArrivalDate(estimatedArrivalDateEntity.getEstimatedArrivalDate()).build();}
}

这里的代码本来是采用了常规的写法，没有采用自定义注解的方式，注解的方式是参考了后面那位大佬的文章，加以修改实现的。因为我的CacheKey可能存在多个属性值的组合。

Annotitions

/*** @author yuanhewei* @date 2024/7/9 14:51* @description*/
@Target(ElementType.METHOD)
@Retention(RetentionPolicy.RUNTIME)
@Documented
public @interface DoubleCache {/*** 缓存名称** @return*/String cacheName();/*** 缓存的key，支持springEL表达式** @return*/String[] key();/*** 过期时间，单位：秒** @return*/long expireTime() default 120;/*** 缓存类型** @return*/CacheType type() default CacheType.FULL;enum CacheType {/*** 存取*/FULL,/*** 只存*/PUT,/*** 删除*/DELETE}
}

Aspect

/*** @author yuanhewei* @date 2024/7/9 14:51* @description*/
@Slf4j
@Component
@Aspect
public class DoubleCacheAspect {@Resourceprivate Cache<String, Object> caffeineCache;@Resourceprivate RedisTemplate<String, Object> redisTemplate;@Pointcut("@annotation(com.itender.redis.annotation.DoubleCache)")public void doubleCachePointcut() {}@Around("doubleCachePointcut()")public Object doAround(ProceedingJoinPoint point) throws Throwable {MethodSignature signature = (MethodSignature) point.getSignature();Method method = signature.getMethod();// 拼接解析springEl表达式的mapString[] paramNames = signature.getParameterNames();Object[] args = point.getArgs();TreeMap<String, Object> treeMap = new TreeMap<>();for (int i = 0; i < paramNames.length; i++) {treeMap.put(paramNames[i], args[i]);}DoubleCache annotation = method.getAnnotation(DoubleCache.class);String elResult = DoubleCacheUtil.arrayParse(Lists.newArrayList(annotation.key()), treeMap);String realKey = annotation.cacheName() + RedisConstants.COLON + elResult;// 强制更新if (annotation.type() == DoubleCache.CacheType.PUT) {Object object = point.proceed();redisTemplate.opsForValue().set(realKey, object, annotation.expireTime(), TimeUnit.SECONDS);caffeineCache.put(realKey, object);return object;}// 删除else if (annotation.type() == DoubleCache.CacheType.DELETE) {redisTemplate.delete(realKey);caffeineCache.invalidate(realKey);return point.proceed();}// 读写，查询CaffeineObject caffeineCacheObj = caffeineCache.getIfPresent(realKey);if (Objects.nonNull(caffeineCacheObj)) {log.info("get data from caffeine");return caffeineCacheObj;}// 查询RedisObject redisCache = redisTemplate.opsForValue().get(realKey);if (Objects.nonNull(redisCache)) {log.info("get data from redis");caffeineCache.put(realKey, redisCache);return redisCache;}log.info("get data from database");Object object = point.proceed();if (Objects.nonNull(object)) {// 写入Redislog.info("get data from database write to cache: {}", object);redisTemplate.opsForValue().set(realKey, object, annotation.expireTime(), TimeUnit.SECONDS);// 写入CaffeinecaffeineCache.put(realKey, object);}return object;}
}

因为注解上的配置要支持Spring的EL表达式。

public static String parse(String elString, SortedMap<String, Object> map) {elString = String.format("#{%s}", elString);// 创建表达式解析器ExpressionParser parser = new SpelExpressionParser();// 通过evaluationContext.setVariable可以在上下文中设定变量。EvaluationContext context = new StandardEvaluationContext();map.forEach(context::setVariable);// 解析表达式Expression expression = parser.parseExpression(elString, new TemplateParserContext());// 使用Expression.getValue()获取表达式的值，这里传入了Evaluation上下文return expression.getValue(context, String.class);}public static String arrayParse(List<String> elStrings, SortedMap<String, Object> map) {List<String> result = Lists.newArrayList();elStrings.forEach(elString -> {elString = String.format("#{%s}", elString);// 创建表达式解析器ExpressionParser parser = new SpelExpressionParser();// 通过evaluationContext.setVariable可以在上下文中设定变量。EvaluationContext context = new StandardEvaluationContext();map.forEach(context::setVariable);// 解析表达式Expression expression = parser.parseExpression(elString, new TemplateParserContext());// 使用Expression.getValue()获取表达式的值，这里传入了Evaluation上下文result.add(expression.getValue(context, String.class));});return String.join(RedisConstants.COLON, result);}

Controller

/*** @author yuanhewei* @date 2024/7/9 14:14* @description*/
@RestController
@RequestMapping("/doubleCache")
public class DoubleCacheController {@Resourceprivate DoubleCacheService doubleCacheService;@PostMapping("/common")public EstimatedArrivalDateEntity getEstimatedArrivalDateCommon(@RequestBody EstimatedArrivalDateEntity estimatedArrivalDate) {return doubleCacheService.getEstimatedArrivalDateCommon(estimatedArrivalDate);}@PostMapping("/annotation")public EstimatedArrivalDateEntity getEstimatedArrivalDate(@RequestBody EstimatedArrivalDateEntity estimatedArrivalDate) {return doubleCacheService.getEstimatedArrivalDate(estimatedArrivalDate);}
}

代码中演示了Redis + Caffeine实现两级缓存的方式，一种是传统常规的方式，另一种是基于注解的方式实现的。具体实现可以根据自己项目中的实际场景。

最后的测试结果也是两种方式都可以实现查询先走一级缓存；一级缓存不存在查询二级缓存，然后写入一级缓存；二级缓存不存在，查询MySQL然后写入二级缓存，再写入一级缓存的目的。测试结果就不贴出来了

总结

本文介绍Redis+Caffeine实现两级缓存的方式。一种是常规的方式，一种的基于注解的方式。具体的实现可根据自己项目中的业务场景。

至于为什么要用Redis+Caffeine的方式，文章也提到了，目前我们Redis集群压力还算挺大的，而且接口对RT的要求也是比较高的。有一点好的就是我们的数据是每天全量推一边，总量也不大，实时性要求也不强。所以就很适合本地缓存的方式。

使用本地缓存也要注意设置容量的大小和过期时间，否则容易出现内存溢出。

其实现实中很多的场景直接使用Redis就可以搞定的，没必要硬要使用Caffeine。这里也只是简单的介绍了最简单基础的实现方式。对于其他一些复杂的场景还要根据自己具体的业务进行设计。我自己也是边学边用。如果有问题或者其他好的实现方式欢迎各位大佬评论，一起进步！！！

参考

https://blog.csdn.net/weixin_45334346/article/details/136310010