Redundancy or trouble? Simultaneous wireless + wired network connections

Is there any harm (or any point) in enabling multiple network interfaces simultaneously on the same system?

Each gets its own IP from the router. Will that cause confusion or does linux / Network Manager handle it?

It’s only confusing if you don’t understand how it works. :wink:

Stick to wired only, keep wireless configured but disabled for “just in case - one day” scenarios.

Care to elaborate?

The idea would be to have a redundancy, specifically if the wireless drops for some reason, especially when I’m not there.

My plan was to use wireless only, since I had previously tested ethernet over powerline in this location and it was unusable. For fun I tried it again now and somehow it works, for now at least. I’m considering keeping it plugged in for redundancy if it doesn’t cause network problems.

I meant ditch the wireless, keep wired … Thinking you actually have regular Ethernet, now I see you’re doing Ethernet over power… Grrr.

Well in that case, keep wireless as primary. If you really want to keep powerline redundancy make sure you set your metrics to a higher number on your secondary powerline interface, and a lower number on your primary.

You’re going to get 2 local ips and 2 default gateways. If you’re doing any automatic port forwarding, you’ll need to make sure it’s using the right interface.

Also, any internet tcp connections will break when your WiFi disconnects but you should be able to re-establish.


IMHO maybe it’s more worth fixing your WiFi?

So, with different priority numbers for the different network interfaces it should avoid problems? Sounds good.

How can I test which interface is being used?


To my knowledge there’s nothing to fix with WiFi between my device and the wireless access point.

But the backhaul between the WAP and router is … yup, ethernet over powerline. On occasion there have been network drops, although the wireless connection didn’t seem to drop. Can’t be sure if it was the client device, the WAP, router, or… the powerline adapters. I’ve long suspected the latter but it’s hard to diagnose.

ip route get also see http://linux-ip.net/html/tools-ip-route.html


Thanks exist things like: https://mikrotik.com/product/sxtsq_5_ac that you can also to build bridges indoors, at least with Mikrotik there’s no shortage of logging

I use between 2 and 3 network connections simultaneously for work. I need access to multiple networks and must prioritize the access for our AD and intranet to stay functional. It works quite well with the exception of some VPN clients which will disable all other interfaces to prevent traffic from traversing networks.
Also some LACP modes use this for failover.

I set priority numbers with network manager (nm-applet) with wifi having a higher number, but it still seems to use the ethernet interface:

$ ip route show
default via 192.168.2.1 dev enp5s0 proto static metric 100 
default via 192.168.2.1 dev wlp6s0 proto static metric 600 
192.168.2.0/24 dev enp5s0 proto kernel scope link src 192.168.2.145 metric 100 
192.168.2.0/24 dev wlp6s0 proto kernel scope link src 192.168.2.144 metric 600

$ ip route get 1.1.1.1
1.1.1.1 via 192.168.2.1 dev enp5s0 src 192.168.2.145 uid 1000 
    cache 

Is there something else I need to do?


Do you mean different network interfaces/devices?

Sounds useful. How is this configured?

Yes, wifi, integrated nic and usb-ethernet adapter.

https://wiki.debian.org/Bonding
Have a look at Example 2. May be of some help to you.

Nice, so it’s called bonding.

Seems ifenslave is deprecated, at least on Fedora. There are instructions on setting up bonding with Network Manager’s command line interface. But some report needing to disable Network Manager for it to work (inconvenient). Then there is something called teaming which works with Network Manager.

Smaller metric == higher priority, it sounds the other way around but that’s how metrics have always worked in networking.

Flip the 100/600 numbers around and try ip route get 1.1.1.1 again.

(Not 100% sure if cache flush would be needed, but probably not as it would make sense for it to be flushed implicitly when changing the routing tables).

That’s what I would have thought too except that in Network Manager, the tooltip says: “Connections with higher numbers are preferred when selecting profiles for automatic activation. Default value is 0.”

I tried again and switched the numbers I used between the profiles (not 100 or 600 btw). Then systemctl restart network.service. Same results.