I’m posting here in case someone ever has a similar issue. I didn’t find any answers when I was searching.
My recorder data store was growing out of control - >80GB in a month. I recently moved it to MariaDB which gave me great ability to query and figure out what was going on.
I have two Roborock vacuums integrated to HA. I use the camera integration for the maps to display.
What I discovered is two-fold: I had over 350K updates from each vacuum, and each update was around 50KB because of all the attribute data. For anyone not keeping track, that’s about 17GB for each vacuum in recorder data. Sure, storage is cheap, but I have no need for that data.
I excluded both camera entities from recorder and ran the recorder purge for those two entities. It took a loooong time to run, but I did end up clearing even more space than I was expecting. Ran a quick table optimization afterwards and my states table went from 40GB to 2GB!
Hope this helps somebody someday!