Home Assistant becomes inaccessible after some time, have to pull the plug every time

I’m running hassOS on a RPI 3A. It seems to be working fine but after some time, can take everything from minutes to hours, it just stops responding. The RPI is stuck with solid red and green lights, I can still ping the RPI but home assistant becomes inaccessible. I can’t use ssh or samba either so I have no way to check the logs. The only way I can get it working again is pulling the plug and putting it back in.

Does anyone know what might make this happen?

Sounds like a dying SD card, and every time you pull the plug it will make it worse.

Make a backup, shift it off the pi, and install to a new Sd card.

1 Like

I’ve read that this could potentially be the problem but also something about using the wrong power supply. Right now I’m using a regular phone charger adapter, could that be the issue?

Could be! You need a consistently good 2-2.5A power supply.

In turn a power problem will screw your SD card, so you may have both problems by now :slight_smile:

1 Like

I good USB power supply which is cheap and available in many countries is this one from IKEA with the name Koppla.

It is dirt cheap. It performs well. And everything I have seen of tear-downs a safe design with proper isolation between mains and the USB outputs.

Equally important is to use a short USB cable of a good quality. And expensive is not always good quality. In fact the best I have found are from another Swedish chain called Biltema. I have measured the voltage drop over many different and they were surprisingly the best.

If I need longer cables I make my own with just two wires for ground and 5V using good wire and USB connectors I buy cheap 50 pcs from China via eBay.

Phone chargers are often only rated for 2.1 A. And often they struggle to keep the voltage when they get near the 2.1 A.


Luckily I live in Sweden so that sounds like great news for me! I suppose you mean this power supply? https://www.ikea.com/se/sv/p/koppla-usb-laddare-med-1-port-vit-40412278/
I would happily have a link to the cable.

Regardless of the source of this problem, I’ve found it very useful to enable watchdog functionality on all of my Linux servers. A hardware watchdog is preferable, but even a software watchdog can be useful if the kernel itself isn’t frozen.

I’ve set up a watchdog on my RPi’s using this guide as a starting point: https://www.meazurem.com/blog/raspberry-pi-with-network-watchdog/

AFAIK the Pi does not have a hardware watchdog chip, although I think an ESP device could be configured to act in this way, cycling power under certain conditions. You’d need to use a relay in line with the Pi’s 5V power source. I have written automations for smart switches to do this for various devices I consider critical in my home - security cam, wifi access points, etc. It’s also easier than climbing up to hard to reach places to do a hard power cycle. :slight_smile:

1 Like

Hmm. It seems they have two with same name

Go for the one with 3 outputs as it us rated for a higher current.

More than 3 A


Be careful with the one with 3 outlets. It gives max 2.4A on one outlet, but only 3.7A in total !

1 Like

Correct. The key is that it can give 2.5 A for the Raspberry. 2A is not enough in practical. But you should only use one (any) of the 3 outputs

1 Like

Phone charger is almost 100% certainty. You need a minimum 2.5Amp for a Pi3… According to the main dev, 95% of this is caused by power. Don’t skimp here.


Thanks for your help, will probably buy the Koppla supply! Did you have a link for the Biltema cable?

No. But it is their grey 1 meter USB to microUSB. Both connectors and wires are grey

So I bought a new class 10 SD card and the power supply and cable recomended here. Same problem appears… I don’t get what is causing it

@MobbareKurtZ, I can’t say for sure that your problem is the exact same as mine was. However, I found that the issue I was running into was NOT that the system was ‘inaccessible’ so much as the cpu was overloaded (100+%). The solution that worked for me was simply to use a MySql server running on another computer for the recorder rather than the builtin recorder. That dropped the cpu usage approx 75-90% and the problem stopped.

Hope this helps.