I’m new to Home Assistant and recently set it up as a VM under Proxmox (Running HAOS under Proxmox).
Home Assistant
Installation methodHome Assistant OS
Core2025.7.2
Supervisor2025.07.1
Operating System16.0
Frontend20250702.2
It keeps crashing and rebooting unexpectedly. The log is filled with repeating errors like:
Error doing job: socket.accept() out of system resource (None)
OSError: [Errno 24] No file descriptors available
Inside the container, the number of open file descriptors is stuck at 1024:
ls /proc/$(pidof python3)/fd | wc -l
1024
Listing the open FDs doesn’t show anything obviously excessive—mostly eventpoll, SQLite files, and individual sockets.
There’s also this warning related to protobuf:
Protobuf gencode version 5.28.1 is exactly one major version older than the runtime version 6.31.1 at pyatv…
I’d really appreciate guidance from the community to confirm what issue I am facing. Is there anyone who can help me understand what’s causing this and how to fix it?
What version of HA are you running?
You can look at the version info page. If it states “Operating System” it is HAOS, if it just states “Supervisor” and not “Operating System” it is Supervised and if it only states “Core” and “Frontend” it is Container or Core.
Turns out the HomeKit Device integration was failing, and after looking into it, one of my Ecobee thermostats kept causing issues, so I disabled it. The TP-Link integration was also failing because one of the smart plugs was no longer reachable—I had replaced it with a Zigbee plug. After removing both, the file descriptor count is now at 439. I’m currently monitoring to see if it remains stable.
I’m running HA 2025.5.3 on ESXi, Supervisor: 2025.07.1, and HASSOS 16.
When I upgraded last month to 2025.6.x, it would immediately crash on reboot with the above error in the logs:
OSError: [Errno 24] No file descriptors available
2025-07-20 02:07:42.144 ERROR (MainThread) [homeassistant] Error doing job: socket.accept() out of system resource (None)
Traceback (most recent call last):
File “/usr/local/lib/python3.13/asyncio/selector_events.py”, line 178, in _accept_connection
conn, addr = sock.accept()
~~~~~~~~~~~^^
File “/usr/local/lib/python3.13/socket.py”, line 295, in accept
fd, addr = self._accept()
~~~~~~~~~~~~^^
I rolled back to 2025.5.3 and left it since I didn’t have time. I did however upgrade the Supervisor and other updates. I just left core. Last night, I woke up to HA being unresponsive. I rebooted and checked the logs, the error above is back… over and over until the box was unresponsive. It had been running for 28+ days though before it occurred so I have no idea why or how to proceed.
I use HomeKit, like OP, and have a few dumb Tuya gas wifi devices. I do not use Frigate, but I do use Zoneminder. If nothing else, I’m posting to say “me too” I guess. Will monitor to see if it occurs again.
Mine didn’t happen since I completely disabled “HomeKit Device” & also when I removed the unreachable TP Link device which was using “TP Link Smart Home” integration.
Not sure TBH, As far as I understood, developers plan is to raise limit to 2048 in one of the next updates. Personally, for the time being, I have rolled back to OS 15.2 where the fds were not limited. It seems like, at least for me, there is an fd leak somewhere : im monitoring the number of used fds and it keeps increasing steadily, not sure which part of my setup is causing this at this point.
Here’s my graph for open fds for the ha container, definitely doesn’t seem healthy
I had just upgraded to OS 16.0 yesterday (also running as a VM under ProxMox) and was hit with this issue as well. It seems like the system was locked up overnight for over 3 hours writing 13 GB to the disk (with errors similar to what was reported here of OSError: [Errno 24] No file descriptors available) and then rebooted itself. All the errors I could see were similar and available RAM was dropping over the course of 9 hours until it eventual ran out of file descriptors. I just reverted back to OS 15.2 and will see if it is back to stable (which is my assumption based on reading this and other threads). Hopefully this will get resolved in a future release.
Oh I’m using datadog (http://www.datadoghq.com/) with a free trial account => I manually installed the datadog agent as a docker container on the home assistant OS, it collects a number of metrics and one of them is the number of open fds per container. Graph is done using the datadog web dashboard ui.
UPD: It was crashing every 2 hours on 16.0, now it is crashing less frequently, stays alive for about 4 hours. But the fact is that it does not crash at all on 15.9 on the same setup.
Same issue for me, rebooting 1 time in 24 hours, OS 16 and Core 2025.8.3, Installed HAOS at VM Proxmox, in other thread said that it solved with increased limit from 1024 to 2048 but problem still exist. I’ve checked this article but still no work.
I have the same problem even upgrading to 16.2. HA web UI couldn’t be access after few hours or minutes. The problematic add-ons were Tapo Camera Control and Dahua Integration. I am downgrading back to 15.2.
