[redis] redis penetration, avalanche and breakdown

through

Definition ：

Penetration means that the requested data is in Redis Neither in the cache nor in the database , So the database returned an empty data ,Redis No caching , Every time the requested data is queried in the database .

resolvent ：

1. Redis Cache the empty data returned by the database , And set a reasonable expiration time , In this way, you can get data from the cache when requesting again .
2. Set up the bloon filter (Bloom Filter), The function is to filter the requested data first , Intercept requests for nonexistent data , The existing data is released to Redis.

An avalanche

Definition ：

Avalanche refers to a large number of Request data ( inequality ) Inquire about Redis when Redis Broken or data expired at the same time , Cause the request to hit the database directly, causing the database to crash .

resolvent ：

1. Set multi-level cache ：nginx cache Redis cache .
2. Set up disaster warning mechanism ： Regular inspection Redis Server status, etc .
3. The data expiration time is set to a fixed value + Random value , Try to spread the expiration time .
4. Current limiting the drop （ Not recommended ）
   • Current limiting ： Limit the number of requests that request the database
   • Downgrade ： For some hot spots, data is saved in memory , When Redis Query memory directly in case of failure
5. Set the hotspot data to never expire
6. Automatically / Manually check the expired data , Decide whether to postpone .

breakdown

Definition ：

Breakdown refers to a large number of requests The same data , But at this time Redis This data is out of date , Cause the request to hit the database directly, causing the database to crash .

resolvent ：

1. Distributed lock , Only one request is allowed to query the database , Other request delay .
2. For hot data settings, never expire .
3. Automatically / Manually check the expired data , Decide whether to postpone .
4. Data preheating , The hot data is saved to the cache in advance .