mDNS (.local) failing randomly

I run HAOS inside proxmox, and alongside it there’s another VM running Ubuntu Server. The latter has avahi-daemon installed so it can be accessed at proxmox-ubuntu-server.local.

Inside Home Assistant, I set up some REST sensors that access http://proxmox-ubuntu-server.local/.... These have been failing randomly, especially since I recently upgraded to Proxmox 9 and HA 2026.1. I’m not sure if it was totally solid before, but it seems less reliable now.

The errors from the REST sensor look like this:

Cannot connect to host proxmox-ubuntu-server.local:80 ssl:default [Connect call failed ('fe80::...:8676', 80, 0, 2)]

To diagnose I also set up a Ping sensor and I’m seeing brief failures on a regular basis (but not an exact interval):

And I also ran nslookup proxmox-ubuntu-server.local in the HA terminal once per second in a loop, and saw these “No answer” errors around the same time that the ping sensor disconnected:

I’m not really sure where to go from here since both of the VMs and the network in general seem healthy. Any tips on what I should try?

Looks like an IPv6 issue. Do you have Ipv6 enabled in your home network? If not you should disable it as well in Proxmox.

It’s correct that ipv6 is not enabled on my router (mikrotik). (It sounded like a bit of a journey to enable it, so I never bothered to do that, which has been fine for pretty much all home network/internet uses so far.)

Is there a way to disable ipv6 for all the VMs at once / via proxmox? I tried adding a conf file to /etc/sysctl.d with net.ipv6.conf.all.disable_ipv6=1. After rebooting, ip a in a proxmox shell no longer shows any inet6 entries. However, in the HAOS VM shell, network info still shows ipv6-related stuff, and ip a inside my Ubuntu Server VM also shows a couple of inet6 addresses. I suppose I could also tweak settings within those VMs, but was hoping there would be a global option of some sort?

Also curious what makes you say this is an ipv6 specific issue – I see the ipv6 address in the “Connect call failed”, but does it also explain why nslookup would fail?

Yes, you probable need to disable it in all VM’s (You can disable it in the HA Network settings). Did you check your /etc/resolv.conf file? In virtualized networking the host often proxies the DNS. If it used a IPv6 address for that this could be a problem.

Also, I would enable IPv6 for everything if your ISP supports it. E.g. Office365 performance is in my experience better over IPv6. Since we ran out of IPv4 addresses IPv6 only sites start to appear.

I enabled IPv6 on my router 14 years ago. Worked fine ever since. Don’t forget that if you have fireballing rules in place, you need them form IPv6 as well.

Yes, mostly I am afraid of not being competent enough to set up a safe firewall. The Mikrotik router OS is already a bit scary complicated for me (mostly bought it for the PoE out :slightly_smiling_face:) and may not have good defaults for ipv6. Maybe when I have some extra free time I will give it another shot.

As far as I know the defaults on router level should be fine (deny all inbound except when connected to an outbound data stream or specific ICMP packets). The main challenge is if you have firewalls within you home infrastructure.

I would start by looking form any information on how your ISP provides IPv6 support and work from there.