HA crashes after +/- 7 day's uptime => no routecase in logging

Installation Home Assistant OS
1

Hi,

I am running Home Assistant 2026.4.3 on an HP i5 machine with MariaDB. HA OS on bare metal,

For the past two months I’ve been experiencing a very strange issue: after about 4–7 days, Home Assistant “crashes”. By this I mean:

  • I can no longer access HA locally (web interface unavailable)
  • The Home Assistant instance completely disappears from my local network (not visible in network browser)

The only way to recover is to manually reboot the server. After reboot, everything works fine again for another 4–7 days.

What I’ve checked so far:

  • Home Assistant Core logs → no relevant errors
  • Supervisor logs → nothing unusual
  • CPU and memory usage → normal, even leading up to the crash
  • System continues running (only HA becomes unreachable)

There are no clear warnings or errors before the crash happens.

Has anyone experienced something similar, or does anyone have suggestions on how to find the root cause of this issue?

Thanks in advance!

2

What chipset is your LAN connection? Ethernet or WiFi connection?

Is the rest of your network affected? Is it your router?

Is it crashed, locked up/unresponsive, or working fine but just unreachable?

Is it a capacity issue? How much RAM, disc space, CPU, both total and used/ free.

Does it follow a pattern? Time of day, after a certain event, etc?

What does your screen on the HP say? Any events or errors, before or after?

You specifically mention MariaDB. Any database issues or doubts about integrity? How big is it?

Any VLANs, or other non-standard setup that might be relevant?

3

Thanks for your reply!

  • LAN: Ethernet (what else :slightly_smiling_face:)
  • Network: No, my network is not affected. Everything in the house that uses Wi-Fi continues to work fine
  • Capacity: After a crash I checked RAM, disk space, and CPU — there are no unusual spikes or strange patterns in the graphs
  • Pattern: Unfortunately, no clear pattern
  • HP machine: I will connect a screen and keyboard to monitor it directly, thanks!
  • MariaDB: I think the database is relatively small, but I still need to verify that
  • VLAN: No, just a plain/standard network setup
4

Did it crash again? (Nearly a week has transpired)

I asked about the network chipset as some are known to have problems, such as some Intel network chips. You overlooked the answer.

The welcome screen often has information that may be relevant. A clear photo can often clarify a lot of issues.

MariaDB: ‘Relatively small’ - give us some actual figures, otherwise we are still grasping at straws and guessing, and you won’t get a solution.

I note an HomeAssistant update has been released since your post that may have solutions to your problem. Be sure to toggle the backup option to ‘on’ before installing it.

5

thanks for asking!!!

Yes, unfortunately it crashed again. I had connected a screen and keyboard, but after the crash both were frozen — no additional information was displayed.

Regarding the network chipset: I have a Realtek RTL8168 (PCI ID 10EC:8168) running the r8169 driver. Apologies for overlooking that question earlier.

Good point about the welcome screen — unfortunately it was completely frozen after the crash, so I couldn't capture any useful information from it.

MariaDB size: 2.6 GB

Claude suggested disabling the hwaccel_args in Frigate, which I've done, and so far no crashes since — but I'm still waiting to see if it holds.

I'll check the latest Home Assistant update and make sure to enable the backup option before installing it. Thanks for the tip.

If the reboots continue I might consider replacing the hardware, since without concrete logs it's likely a hardware issue. I've had my HP ProDesk 400 G4 Mini for 2 years (second-hand) and I'm happy with its low power consumption — but this instability is really frustrating.

6

Please, I would like to follow the progress of your case. I'm noticing a lot of similarity between your case and mine, however, I use Proxmox. On it I have HA, Frigate, Z2M and other CTs. I've been experiencing frequent crashes due to OOM or SWAP. On Frigate I have 12 cameras in an LXC with 8 cores and 10GB of RAM. Recently I started using GENAI and it increased memory usage for embeddings. However, I've always noticed a memory leak behavior in the Frigate's LXC, and this already happened when I ran it directly in HA. Yesterday, when my last crash occurred, following Gemini's advice, I remounted the swap disk, but I also disabled GENAI on Frigate. Embeddings usage is lower and I'm monitoring the system behavior. Thank you for any information on the progress. I've kept hwaccel_args for now.

1 Like
7

Wow, you have an impressive system. Sorry to hear that you’re experiencing crashes as well. I’ll report my progress in this thread. I used Proxmox too, but a couple of months ago I moved to Home Assistant OS to keep the system as clean and simple as possible.

1 Like