supervisor: Latest supervisor update broken networking

Describe the issue you are experiencing

Using a netbooted diskless raspberry pi4 system running a supervised-install - https://github.com/home-assistant/supervised-installer - like that.

As of today the system loses it’s IP address shortly after starting the supervisor and will never recover.

I cannot reach the web UI to pull version information - it is up to date.

What is the used version of the Supervisor?

2022.08.6

What type of installation are you running?

Home Assistant Supervised

Which operating system are you running on?

Debian

What is the version of your installed operating system?

11.4

What version of Home Assistant Core is installed?

2022.9.0

Steps to reproduce the issue

1.boot the box 2.wait 3.fail

Anything in the Supervisor logs that might be useful for us?

Unable to access the logs

Additional information

This setup has been working for years - like home assistant 0.66 or something like that. I have no idea why a booted working system having it’s network checked causes it to go offline but based on the changes being network related I’m blaming these changes.

https://github.com/home-assistant/supervisor/pull/3828 https://github.com/home-assistant/supervisor/pull/3829 https://github.com/home-assistant/supervisor/pull/3830

It does effect multiple systems I have running this config.

About this issue

  • Original URL
  • State: closed
  • Created 2 years ago
  • Reactions: 5
  • Comments: 28 (4 by maintainers)

Most upvoted comments

$ /sbin/ifconfig eth0:0
eth0:0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>  mtu 1500
        inet 192.169.99.41  netmask 255.255.255.0  broadcast 192.169.99.255
        ether dc:a6:32:c3:b5:19  txqueuelen 1000  (Ethernet)

was checking to see if the supervisor had started…

35769eba5f2   ghcr.io/home-assistant/armv7-hassio-supervisor:latest    "/init"   8 days ago   Up 9 seconds                                            hassio_supervisor

I ran another docker ps just after that and it locked up.

It’s too bad cause that would have been a nice easy workaround. For now I’ll keep using the eth0 + wlan0 config that works.

heh - this is printing to console… 🙄

nfs: server 192.168.99.25 not responding, still trying

Oh I do like the let HA have the alias idea… will try it and post back.

Well that change very nearly ruined my Saturday. Thank goodness for the USB NIC… bacon saver for sure.

There is one other minor mod I have made - not sure when - done quite some time ago after I noticed my system would empty it’s /etc/resolv.conf file shortly after booting up with a helpful line at the top…

# Generated by NetworkManager So at least I knew who to blame.

~$ cat /etc/NetworkManager/NetworkManager.conf
[main]
#dns=default
dns=none

another leave my network alone setting

Anyway… what I learned from looking into this is that the supervisor uses NetworkManager, so I came up with another workaround, which is to tell NetworkManager to leave the interface alone: edit /etc/NetworkManager/NetworkManager.conf by adding ;type:ethernet to unmanaged-devices=type:bridge;type:tun;driver:veth under section [keyfile] That should make the interface go from connected to unmanaged in nmcli dev status (requires a reboot; restarting NM would be enough, but that kills my system aswell)

@Mofef just a heads up, this may have other impacts. Supervisor and addons essentially only acknowledge the existence of managed interfaces. Which explains why it works (Supervisor ignores it entirely) but also has other consequences. For example:

  1. Cannot update network info from HA UI or CLI
  2. Cannot use the Samba addon
  3. Cannot use the Adguard addon
  4. …Probably other addons too, those are the ones I remember from issues

Honestly probably other stuff too I’m forgetting. HA will basically think you have no network interfaces.

Seriously though - @mdegat01 thanks for looking at this.

Very glad to hear I’m not the only lunatic running a setup like this though @Traviis 🍻 gooooo diskless!! 🤪