Moet of those entties are dead weight and are no longer in use and never been relevant. Been using Bermuda en Shelly plugs to do presence detection in key rooms. That part works great, except now I have a metrics fuckton of useless entities that are loaded every time we open the app/WebUI. Nuba Casa seems to me capped at 700KB/s so having to load 170K entities sized at 28MB takes about 40 seconds. Tried calling purge entities with device_tracker.*_*_*_*_*_* (and various variations) to no avail. Starting up HA takes several minutes and it's memory is hovering just above 2GB. Getting pretty desperate figuring out a fix besides disabling everything.
#Access HA through Nuba Casa is incredibly slow due to 170K device tracker entities
lost monolith
worn barn
This isn’t really a Nabu casa issue. Did you get rid of known_devices in config and restart?
lost monolith
Just deleted it now and starting HA back up.
Not entirely agreeing it's not a Nuba Casa issue, but the fix for sure isn't there.
Holy shit thanks @worn barn, it loads instantly and the UI is super snappy again.
Where could I have found this in the docs? Kinda assuming I missed it somewhere
Now at 600MB instead of 2GB 🎉
worn barn
It will be recreated and re-populated. So over time you might run into this again. Using the BT device tracking is something that requires some kind of filtering I imagine. It sounds like this was bogging down the system
lost monolith
Ok so the bug still persists then, lets see if I can prevent it from beeing created
Yeah I know it was bogging it down and slowly grinded it to a halt
worn barn
The connection is handled by HA, and not the host system running it, so anything increasing resource usage within HA can have an effect on the NC connection.
What I mean by that is that it is not like a remote desktop app client such as teamviewer etc running independently
lost monolith
It was mainly slowing the UIs down. When no one was looking at it it stayed around 20% CPU usage. Even when opening the UI it stayed around that. But starting up takes a long time. So aside from a few specific moments HA is self was fine.
Opening the WebUI from the local network was blazing fast and didn't have the same bandwidth limit as through NC.
90 seconds till HA is reachable through NC now 🎉