How to turn off DB cache completely


We’ve got some interesting issue with kong (1.1.1) and cassandra on k8s.
Due to some race condition (or rather outdated cache) there is a possibility to add a route to the service that has just been deleted

We are checking db_update_frequency but it doesn’t seem to turn off the DB cache completely.
There is the db_cache_ttl parameter, but setting it to 0 turns off the expiration of that cache rather than turning off the cache at all.

The question is:
How to turn off DB cache completely, or at least make it as volatile as possible?