Cache

Purpose

Due to the large database read operations during the master-server scheduling process. Such as read tables like tenant, user, processDefinition, etc. Operations stress read pressure to the DB, and slow down the entire core scheduling process.

By considering this part of the business data is a high-read and low-write scenario, a cache module is introduced to reduce the DB read pressure and speed up the core scheduling process.

Cache Settings

  1. spring:
  2.   cache:
  3.     # default disable cache, you can enable by `type: caffeine`
  4.     type: none
  5.     cache-names:
  6.       - tenant
  7.       - user
  8.       - processDefinition
  9.       - processTaskRelation
  10.       - taskDefinition
  11.     caffeine:
  12.       spec: maximumSize=100,expireAfterWrite=300s,recordStats

The cache module uses spring-cache, so you can set cache config like whether to enable cache (none to disable by default), cache types in the spring application.yaml directly.

Currently, implements the config of caffeine, you can assign cache configs like cache size, expire time, etc.

Cache Read

The cache module adopts the @Cacheable annotation from spring-cache and you can annotate the annotation in the related mapper layer. Refer to the TenantMapper.

Cache Evict

The business data updates come from the api-server, and the cache side is in the master-server. Then it is necessary to monitor the data updates from the api-server (use aspect point cut interceptor @CacheEvict), and notify the master-server of cacheEvictCommand when processing a cache eviction.

Note: the final strategy for cache update comes from the expiration strategy configuration in caffeine, therefore configure it under the business scenarios;

The sequence diagram shows below:

cache-evict