#After 2024.7: Recorder Purge fails and all Sensors consequently stop recording!
austere egret
I‘m on HAOS on an x86 machine
It started with the 2024.7 update
After 2024.7: All Sensors stop updating after Samba Backup starts
austere egret
Hmm, i actually can't exactly tell if the problem is Samba Backup, it starts creating the backup at 4:00am in the morning, and starts copying the backup at 4:25am, however all sensors stop recording at 4:12am
This is such a weird behavior. I've now adjusted the samba start time so I can tell if it's causality or correlation ...
if anybody has any troubleshooting tips, I'd be very happy for any hints!
I just rechecked the logs, and there isn't even a log-line for the time frame in question 😦
it weird, really all sensors stop recording, from systemmonitor via mqtt to esphome and others.
3 days in the row, same time of the day, only fixed by restarting
austere egret
Oh this indeed sounds similar, thank you so much!
yes seems the problem is the recorder purge failing. I'm not entirely clear yet what causes it, I'll check this again later. Hopefully this will see a hotfix, as this is quite a large break in the system! Lost almost a full day of data recording because I didn't notice it until rather late in the day 😦
After 2024.7: Recorder Purge fails and all Sensors consequently stop recording!
actually just found this in the logs:
2024-07-06 06:52:48.930 ERROR (MainThread) [homeassistant.components.recorder.core] The recorder backlog queue reached the maximum size of 249278 events; usually, the system is CPU bound, I/O bound, or the database is corrupt due to a disk problem; The recorder will stop recording events to avoid running out of memory
austere egret
I temporarily fixed it by setting auto_purge: false in my recorder settings in my configuration.yaml. There is a fix that is scheduled for 2024.8 which I think runs very late.
arctic nacelle
From what I read: Because of SQLite limitations it’s a very very invasive fix. Bug has existed since may. Unclear why it’s suddenly gotten worse, but a safer workaround hasn’t been found yet. Unclear why some people have the broken database index and some don’t. Maybe a custom integration did something dodgey.
