In my UNRAID network setup, I have defined three DNS servers:

1) My own - Which is a technetium docker within UNRAID, accessible fine by everybody (other PC use it fine for resolution and I can access its web interface), set in its own IP, as the docker is defined to use "br0" network.
2) My router.

3) 8.8.8.8 (as a fall back)

 

Seeing UNRAID's /etc/resolv.conf indeed shows those three IP.

When I try to ping addresses that are defined in my internal DNS, it doesn't resolve them. Using nslookup, I find that it uses my router as DNS!

What is wrong?

 

Note that from within UNRAID console my own DNS does NOT respond to ping, while from my PC it does.

UNRAID cannot access its own docker network?

(I verified with a second docket that uses br0 and own IP... pingable from my PC, not pingable from UNRAID that hosts it)

Edited February 28, 20251 yr by NLS

53 minutes ago, NLS said:

 

UNRAID cannot access its own docker network?

(I verified with a second docket that uses br0 and own IP... pingable from my PC, not pingable from UNRAID that hosts it)

Thats a secruity feature by docker.

You can enable host access to custom networks in the docker settings to work around it. (could have issues with wireguard, if you use that)

2 minutes ago, Mainfrezzer said:

Thats a secruity feature by docker.

You can enable host access to custom networks in the docker settings to work around it. (could have issues with wireguard, if you use that)

 

Can you point me to how?

Also can you point me on any documentation of why this would be a problem with wireguard? (my router does have it's own VPN features, including wireguard, so I might not use this UNRAID feature).

 

BTW I might (and probably have) always had this issue without knowing it. My previous router, was able to manage distinct DNS entries depending on domain (a feature I was using), so probably was used by UNRAID but then transparently still used my own DNS - because my router queried it instead of UNRAID. So I didn't see the problem. The current router (as stupid AVM Fritz by my ISP), cannot do that, cannot even set two DNS in its own DHCP.

 

...ah I found it.

 

Edited February 28, 20251 yr by NLS

Just now, NLS said:

 

Can you point me to how?

Also can you point me on any documentation of why this would be a problem with wireguard? (my router does have it's own VPN features, including wireguard, so I might not use this UNRAID feature).

 

BTW I might (and probably have) always had this issue without knowing it. My previous router, was able to manage distinct DNS entries depending on domain (a feature I was using), so probably was used by UNRAID but then transparently still used my own DNS - because my router queried it instead of UNRAID. So I didn't see the problem. The current router (as stupid AVM Fritz by my ISP), cannot do that, cannot even set two DNS in its own DHCP.

 

Click in the GUI on Settings then Docker 

Stop the Docker Service and enable Advance view in the top right


There is no documentation about the wireguard issue, apart from a thread by me to work around it.

What it boils down to is that unraid uses br0 for interfaces, which will no longer work if host access is activated because it should use shim-br0

Can you point me to the thread to read it? Thanks for the help.

 

Just now, NLS said:

Can you point me to the thread to read it? Thanks for the help.

 

Absolutely
 

 

A note that this also resolved my fetchmail container issue.

(apparently it couldn't find my mail server VM using its FQDN, as it was defined in the DNS it wasn't using)

 

I want to re-open this thread.
Since Monday, I again cannot use my fetchmail container.
My Technitium DNS is working (it uses its own separate IP), my PC is resolving things using it.

BUT again fetchmail cannot ping it (container-to-container), although "host access to custom networks" is still enabled since last time (that worked fine).

Weirdly enough UNRAID itself (as host) ALSO cannot ping my DNS. My PC pings my DNS.

I am in 7.1.2 UNRAID, but I am since way more than the beginning of the week. I feel something else changed. Maybe some plugin update? Could it be something related to tailscale? (which I don't actively use but have installed)

Any ideas?

...and I fixed it again.

Seems my docker network was using ipvlan... I switched to macvlan and it worked...

Is this ok to leave it like that?

(also I disabled, applied and re-enabled "host access" setting again... could it be this that fixed it?)