What evidence do you have for this? This seems impossible to me. HA is what tells the DB to add something. If it crashed, what would tell the DB to add a row? There’s essentially only 3 things recorded in the DB from HA:
- Events fired over the bus (exception:
state_changedevents since state changes are different) - Changes in state for entities (including attributes)
- Statistics for entities with numeric states
Almost certainly if your DB is increasingly rapidly you have some integration that is rapid firing state changes at HA. See here for some tips on finding those entities.
That’s not how this works. Repack isn’t an HA thing, its a native database thing. It does not remove any actual user data. That would be a major, major issue because admins do repacks on production databases regularly. It essentially does a clean up, re-organizing data on disk to optimize indices and layout and such. It also does admin things like cleaning up the transaction table (which may remove some data but not user data, internal data). Among other stuff, I don’t really know all it does, I’m not a db expert either.
But the point is it is something HA is triggering in the DB, it is not a process written by HA. Therefore it has absolutely no knowledge of what you have included and excluded in your recorder config, it doesn’t even know what is in the tables. Everything is just bytes to it, it would never delete your stuff.