Statistics are deleted every night

Configuration
,
1

Hi all,

since a few days I have the issue that my Home Assistant OS running on Proxmox (installed using the formerly tteck community scripts) decides to delete all statistics somewhere around 4am in the morning. I’ve been running HAOS for almost a year, until now everything went fine.
Does somebody know this behaviour, any pointers as to what could cause this? For now I’m trying to restore a backup every morning to not loose everything, but restoration process seems a bit dodgy as well.
Would love any help! I couldn’t figure out a better Category, let me know if I did something wrong. I’ll provide any missing information asap if I can :slight_smile:

Cheers!
Andi

2

That is when Home Assistant performs its database maintenance.

Have you set any custom recorder settings (excludes, includes, purge_kepp_days, etc…)

What do your logs say at around that time (settings → system → logs)?

1 Like
3

Hi Tom,
thanks for your reply!
I have no specific recorder settings configured, there is no “recorder:” line in my configuration.yaml. Everything has been working for quite some time and I cannot remember changing anything, especially not in the configuration.yaml.
For logs I assume I should check Home Assistant Core, right? I did the backup restoration, so I cannot check for today, but I assume that it will happen again tomorrow and I’ll check then and get back to you.
Thanks for helping! :slight_smile:

4

that’s normal. You are using default_config. Which includes it.

Correct.

1 Like
5

Yeah, should have checked logs sooner - seems my database is corrupted:

Logger: homeassistant.helpers.recorder
Quelle: helpers/recorder.py:95
Erstmals aufgetreten: 04:12:35 (1 Vorkommnis)
Zuletzt protokolliert: 04:12:35

Error executing query

Logger: homeassistant.components.recorder.core
Quelle: components/recorder/core.py:907
Integration: Recorder
Erstmals aufgetreten: 04:12:35 (1 Vorkommnis)
Zuletzt protokolliert: 04:12:35

Unrecoverable sqlite3 database corruption detected: (sqlite3.DatabaseError) database disk image is malformed [SQL: UPDATE states SET old_state_id=? WHERE states.old_state_id IN

Logger: homeassistant.components.recorder.util
Quelle: components/recorder/util.py:308
Integration: Recorder (Dokumentation, Probleme)
Erstmals aufgetreten: 04:12:35 (1 Vorkommnis)
Zuletzt protokolliert: 04:12:35

The system will rename the corrupt database file //config/home-assistant_v2.db to //config/home-assistant_v2.db.corrupt.2025-10-21T02:12:35.414362+00:00 in order to allow startup to proceed

Can I do something about that corruption? I guess that would be more of a SQlite-question, but maybe that came up already. I store my important sensor readings in an InfluxDB, so not thaaat big of a problem, but I’d like to use the energy dashboard in HA, which is now completely empty :smiley:

Thanks for all the support! :slight_smile:

Full error messages here: Database Error HA - Pastes.io
Pastebin flagged it as somehow offensive, don’t know what that’s about, if you prefer any other site for sharing let me know.

6

Given you’re storing the important data elsewhere. The easiest thing to do is to stop HA, delete that config/home-assistant_v2.db file & restart HA.

A new dB file with no corrupt entries will be created.

7

I’m getting the feeling this will be the way to go, but this will break my energy dashboard in HA… which is a shame, since I like that very much.
So still in the market for fixes, if there is no other choice I guess I’ll need to make the clean cut… :man_shrugging:

8

Search the forum for database repair topics. Some people have had success in the past.

1 Like
9

Okay, short feedback - I tried a lot of stuff, nothing worked in the end. I dropped the table of the database that seemed to be corrupted, short_term_statistics, but that broke the energy dashboard as well, so in the end I deleted everything and switched over to MariaDB. Hopefully that will make things a little bit more robust. I also limited the number of tracked entities, keep fingers crossed for the future.

But it’s very annoying that there is no possibility to fix something like that, we have very nice things like the energy dashboard, but the value of that greatly diminishes if I cannot compare for example years. I would like the database to be more robust, but maybe that was a mistake or misconfiguration on my side.

Learned a lot, thanks for the help!