Home Assistant 2026.9.1: frontend inaccessible after restart, port 8123 bound to 127.0.0.1

Hi everyone,

Since upgrading to Home Assistant Core 2026.9.1, my frontend becomes inaccessible from the local network after restarting Home Assistant Core.

My setup:

  • Raspberry Pi 4
  • Home Assistant OS 18.2
  • Home Assistant Core 2026.9.1
  • Supervisor 2026.09.0
  • aarch64
  • HA IP address: 192.168.1.114

After a normal restart:

$ ss -lntp | grep 8123
LISTEN 0 128 127.0.0.1:8123 0.0.0.0:*

The frontend is therefore only available from localhost and cannot be accessed at http://192.168.1.114:8123.

The Core logs show:

Error handling when_setup callback for frontend

File ".../components/http/server.py", line 548, in start
    await self._server.start_serving()

OSError: [Errno 98] Address in use

What is strange is that the HTTP configuration explicitly contains:

"server_port": 8123,
"server_host": [
  "0.0.0.0",
  "127.0.0.1",
  "::1"
]

I also verified that no other process is using port 8123 before Home Assistant starts:

ha core stop
ss -lntp | grep 8123

→ no result.

Then:

ha core start
ss -lntp | grep 8123

→ 127.0.0.1:8123 appears.

I also temporarily removed the entire /config/custom_components directory. The problem remained unchanged.

At one point, restarting Home Assistant in Safe Mode resulted in:

0.0.0.0:8123

and the frontend was then accessible from the LAN. However, subsequent Safe Mode restarts reproduced the localhost-only behaviour, so Safe Mode is not currently a reliable workaround.

The network itself is working correctly. When Home Assistant does listen on 0.0.0.0:8123, the frontend is immediately accessible from the LAN.

Has anyone else experienced this behaviour with Core 2026.9.x?

I am also opening a GitHub issue with the detailed logs and troubleshooting results (Core 2026.9.1 binds HTTP server to 127.0.0.1 and fails with EADDRINUSE on port 8123 · Issue #181870 · home-assistant/core · GitHub).

Thanks!

Update — confirmed regression

I have now reproduced and isolated the issue by downgrading the exact same Home Assistant installation from Core 2026.9.1 to Core 2026.8.3.

With Core 2026.9.1:

127.0.0.1:8123
[::1]:8123

and the Core log contains:

OSError: [Errno 98] Address in use

After stopping Core, port 8123 is completely free. As soon as Core 2026.9.1 starts, the localhost listeners appear.

I then downgraded the same installation to Core 2026.8.3, without restoring a backup or changing the Home Assistant configuration.

Immediately after starting 2026.8.3:

0.0.0.0:8123
[::1]:8123

and:

curl -I http://192.168.1.114:8123/
HTTP/1.1 405 Method Not Allowed

The frontend is accessible from the LAN again.

So this now appears to be a Core 2026.9.1 regression in HTTP server/listen-address handling, rather than a network, DNS, configuration, or custom integration issue.

I have also opened GitHub issue #181870 with the detailed diagnostics.

For completeness, I have a separate DNS resolver issue reported as #181849, but the 8123/frontend problem appears to be independent: reverting only the Core version to 2026.8.3 immediately restores normal HTTP binding.