for example, if several cluster are enabled, it is impossible for each cluster to instantiate its own database.
if not, then master instantiates the database, and cluster requests master if it needs to query data.
as a result, it feels like the amount of code has increased a lot.
how did you solve it?