Github user manishgupta88 commented on the issue:
https://github.com/apache/carbondata/pull/2778
@xuchuanyin .....`CarbonLRUCache` instance is one per JVM but cache
implementation is different for different purpose like we have separate cache
provider implementation for blockDataMap and dictionary. You can check the
`CacheProvider` class to get some more details on it.
Cache is a standard interface which has very generic methods. Because we
are storing all the entries in LRU cache to control the memory through LRU
eviction all the keys are being stored in LRU cache map. BloomDataMap should
also be aware about its keys and then use the invalidate API to clean all the
required keys. Check the `clear` method implementation of BlockletDataMpaFactory
---
