Avalanche cache, the cache penetrate because of the cache breakdown problems and solutions

Cache avalanche

Occurred during

Suppose there follows a system, the peak request 5000 times / second, 4,000 times gone cache, only 1,000 times fell on the database, the database concurrency 1000 per second is a normal indicator can work, but if the cache dang machine, or set the same cache expiration time, leading to failure of the cache while at the same time , 5,000 requests per second will all fall on the database, the database immediately is dead, because one second anti database request up to 2000, If the DBA to restart the database, it will be killed immediately a new request, which is cached avalanche.

Solution

1. Beforehand: redis availability, master-slave + Sentinel, redis cluster, avoiding total collapse
2. Things in: ehcache local cache + hystrix limiting & downgrade to avoid being killed by MySQL
3. Afterwards: redis persistence RDB + AOF, quickly recover cache data
4. Cache expiration time is set to a random value, while avoiding failure

Cache penetration

Occurred during

If the second client sends a request 5000, where 4000 is the malicious hacker attacks, i.e., you can not find in the database. For example, the user id is a positive number, user id hackers configuration is negative, if it has been sent 4000 requests per second, the hacker, the cache does not work, the database quickly killed.

Solution

1. Checking the request parameters, return directly unreasonable
2. Also can not find the data into the cache, value is empty, as set -999 ""
3. Use Bloom filter, to quickly determine whether a key is present in the database, there is no direct return

Cache breakdown

Occurred during

Set an expiration time of the key, carrying high concurrency, a hotspot data. From this key expired some time to reload the data into the cache from MySQL, a large number of requests it is possible to kill the database. Refers to a large number of cache avalanche cache miss, the cache means caches hot data breakdown failure

Solution

1. Set key never expires, or when to expire, resetting the key by another asynchronous thread
2. When the data from the cache to get is null, re-lock the database to load data from the process, here to write a distributed lock to achieve demo

Redis achieve Distributed Lock

Previous article I wrote about the principle Redis implementation of distributed locks, there is no longer a detailed overview

Redis distributed lock Why write?

1. Run the lock

SET resource_name random_value NX PX 30000

2. Unlock script execution

if redis.call("get", KEYS[1]) == ARGV[1] then 
    return redis.call("del", KEYS[1]) 
else 
    return 0 
end

Write a distributed lock tools

public class LockUtil {

    private static final String OK = "OK";
    private static final Long LONG_ONE = 1L;
    private static final String script = "if redis.call('get', KEYS[1]) == ARGV[1] then return redis.call('del', KEYS[1]) else return 0 end";

    public static boolean tryLock(String key, String value, long expire) {
        Jedis jedis = RedisPool.getJedis();
        SetParams setParams = new SetParams();
        setParams.nx().px(expire);
        return OK.equals(jedis.set(key, value, setParams));
    }

    public static boolean releaseLock(String key, String value) {
        Jedis jedis = RedisPool.getJedis();
        return LONG_ONE.equals(jedis.eval(script, 1, key, value));
    }
}

Tools to write quite simple

Sample Code

public String getData(String key) {
	String lockKey = "key";
	String lockValue = String.valueOf(System.currentTimeMillis());
	long expireTime = 1000L;
	String value = getFromRedis(key);
	if (value == null) {
		if (LockUtil.tryLock(lockKey, lockValue, expireTime)) {
			// 从数据库取值并放到redis中
			LockUtil.releaseLock(lockKey, lockValue);
		} else {
			// sleep一段时间再从缓存中拿
			Thread.sleep(100);
			getFromRedis(key);
		}
	}
	return value;
}

Reference blog

[1]https://blog.csdn.net/zeb_perfect/article/details/54135506
[2]https://blog.csdn.net/kongtiao5/article/details/82771694
[3]https://www.javashitang.com/?p=596