Redis lost key and bigkey

lose key （ Cache pollution

• The phenomenon ： There are always redis-key Lost situation , Yesterday's key Not today
• Investigation angle ：
  • 1、key Whether to set the expiration time
  • 2、 Check key Delete logic
  • 3、 see redis Whether the memory used in is close to maxmemory, Causes the memory obsolescence policy to be triggered
• If this is the third case , Set according to the business scenario config
  • modify redis-config
    • modify maxmemory
    • modify maxmemory-policy from allkeys-lru Change it to volatile-lru
  • See what key It takes up a lot of memory (bigkeys screening ）, And see if the timeout is set

bigkeys

bigkeys：hash/set/zset There are a lot of elements in ,string The data stored is very large

• harm ：
  • 1、 Blocking 、redis It's single threaded , Yes bigkeys The operation of will cause blocking of other commands , Lead to a large number of redis A series of problems such as operation exceeding and reporting errors
  • 2、 Be overdue / When deleted, it will block , If it is redis4.0 Later versions , Asynchronous deletion can be used to solve （lazy-free/unlink)
  • 3、 Memory is unevenly distributed , For example redis-cluster It will cause uneven use of node memory , Increase the processing pressure of a single instance
• screening bigkeys Methods ：
  • 1、redis-cli --bigkeys
  • 2、 Use open source tools ,redis-rdb-tools To analyze redis Of key Distribution situation https://github.com/sripathikrishnan/redis-rdb-tools

rdb --command json dump.rdb

• resolvent ：
  • Optimize storage solutions
    • bigkey Split , Fragmentation storage
    • Save to database
    • Set expiration time, etc