Can't reach GitHub (and other sites) when Home Assistant is running

Installation
1

Hi,
I have installed Home Assistant on an Optiplex running a fresh installation of Ubuntu Server. When Home Assistant is running, I can’t access some websites like GitHub even from a terminal that has nothing to do with Home Assisant. As soon as I stop Home Assistant, it works fine.

For example, when I do ping github.com, I get the response

ping: Do you want to ping broadcast? Then -b. If not, check your local firewall rules.

I also ran ip route get $(getent hosts github.com | awk '{print $1}' | head -n1) which gave me

140.82.121.4 via my-gateway fdev enp3s0 src my-ip uid 1000
    cache

when Home Assistant was not running and

broadcast 140.82.121.4 via my-gateway dev enp3s0 src my-ip uid 1000
    cache <local,brd>

when it was running.

This happens for a few pages, with GitHub being the most prominent.
I also have a second machine with the same Ubuntu version, that runs the same docker compose and works perfectly fine.

Any help would be greatly appreciated.

2

I tested around a little more and this only happens, when the machine tries to connect to a IPv4 adress. IPv6 works fine.

3

What Ubuntu Version/Release did you install? What Home Assistant installation do you use? Do you see messages in the log “Network unreachable”?

4

I’m on Ubuntu 24.04.3 LTS and have installed Home Assistant Container via Portainer. The Home Assistant version is 2026.2.2.
In the logs I found this:

2026-02-19 16:42:18.860 ERROR (MainThread) [metno] Access to https://aa015h6buqvih86i1.api.met.no/weatherapi/locationforecast/2.0/complete returned error ‘ClientConnectorError’
2026-02-19 16:42:19.533 ERROR (MainThread) [metno] Access to https://aa015h6buqvih86i1.api.met.no/weatherapi/locationforecast/2.0/complete returned error ‘ClientConnectorError’

Is that what you meant?

5

Not exactly, but i’m getting errors with “Network unreachable” with a similar configuration on an Odroid M1 and i’m trying to find a solution for some time already.
I have some some docker containers running and my observation is, when i stop one of these containers the error disappears until i retart home assistant.
What linux kernel version are you using, “uname -a”?

6

I’m on 6.8.0-100-generic

7

@RaphaelRatz
This info might help…
Can you copy / return this? Link to System info – My Home Assistant

8

I have redacted the IP adresses, but they are the ones I see in ip a for enp3s0.

System Information

version core-2026.2.2
installation_type Home Assistant Container
dev false
hassio false
docker true
container_arch amd64
user root
virtualenv false
python_version 3.13.11
os_name Linux
os_version 6.8.0-100-generic
arch x86_64
timezone Europe/Berlin
config_dir /config
Home Assistant Cloud
logged_in false
can_reach_cert_server ok
can_reach_cloud_auth ok
can_reach_cloud failed to load: unreachable
Dashboards
dashboards 2
resources 0
views 0
mode storage
Network Configuration
adapters lo (disabled), enp3s0 (enabled, default, auto), docker0 (disabled), vethb45c2d7 (disabled)
ipv4_addresses lo (127.0.0.1/8), enp3s0 (192.168.xxx.xxx/24), docker0 (172.17.0.1/16), vethb45c2d7 ()
ipv6_addresses lo (::1/128), enp3s0 (2003:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx/64, fe80::56bf:64ff:fe92:1cfe/64), docker0 (fe80::2cb8:81ff:fe63:30ba/64), vethb45c2d7 (fe80::8062:64ff:fe6a:b976/64)
announce_addresses 192.168.xxx.xxx, 2003:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:xxxx, fe80::56bf:64ff:fe92:1cfe
Recorder
oldest_recorder_run February 20, 2026 at 8:11 AM
current_recorder_run February 20, 2026 at 9:11 AM
estimated_db_size 0.19 MiB
database_engine sqlite
database_version 3.49.2
1 Like
9

I’ve figured out a workaround: If I run sudo ip route flush cache after starting the container, it works. Not perfect, but good enough for me.

10

That also seems to help in my case. I will use this as a starting point for further investigation.

11

You could have an eye on this issue. It might be related to your bug and system constellation: https://github.com/home-assistant/core/issues/162636

13

For local ones, they are NOT hackable, so no need, however as you have masked them, the critical information needed has been suppressed.
The symptoms are consistent with two devices using the same IP Address, possibly by two copies of docker containers.
Show the local addresses - the vital clues.
In any case, you appear to have a network configuration error, hence your data packets are not connecting to the outside world properly. Your best option is to go back and follow the configuration documentation, the one supplied by the vendor, very closely. Using guides from ChatGPT, AI and YouBoob sources will most likely confuse you with outdated information.