The Cheap Dongle That Cost Me a Weekend

When we left off in Part 1, I had a router with one Ethernet port and a router operating system that politely insists on two. The grown-up solution was to buy a proper dual-NIC box. The solution I actually went with was an older TP-Link USB 2.0 Ethernet dongle for the price of a sandwich.

I’d like to tell you I agonised over this decision. I did not. I saw “adds a second Ethernet port,” I saw the price, and I clicked Buy with the serene confidence of a man about to learn something the hard way.

The dream of the £10 second port

The pitch is irresistible. Your router box is one port short. Here is a little adapter that plugs into a spare USB socket and gives you another Ethernet port. It’s cheap, it’s instant, and it sidesteps the whole “I need new hardware” problem.

On paper it’s perfect. In practice, USB networking and router operating systems have a relationship best described as strained.

The dongle arrived. I plugged it in. OPNsense saw it. For about a day, I was a genius.

Where it all went sideways

Three separate problems showed up, and they showed up in the worst possible order — slowly, then all at once.

Problem one: the wrong generation. I’d bought purely on price, and the price-shaped clue I ignored was the “2.0” in the name. It was a USB 2.0 adapter, not USB 3.0 — a full generation behind. USB 2.0 tops out at 480 Mbps in theory and noticeably less in the real world, which means it physically cannot carry a full gigabit no matter how good your internet or your network is. My shiny gigabit LAN was being quietly throttled by a sandwich-priced dongle — the network equivalent of putting a garden hose on a fire hydrant. Buying on price alone meant I never checked the one spec that mattered.

Problem two: reliability. This is the big one, and it’s not just me being unlucky. USB Ethernet is widely considered an unreliable option for OPNsense and pfSense, and Realtek-based adapters in particular are notorious for it. The community forums are full of the same symptoms I hit: IO errors, the adapter vanishing after a reboot, the interface simply not coming back. (OPNsense forum thread, FreeBSD forums.)

Why does this happen? Router operating systems like OPNsense are built on FreeBSD, and FreeBSD’s driver support for USB network chipsets is patchy. A cheap Realtek USB adapter that runs flawlessly on Windows can be a flaky mess on a firewall OS. The chip works; the driver doesn’t trust it.

Problem three — the one that actually killed the idea: no VLAN tagging. Remember from Part 1 that my whole single-port survival plan depended on router-on-a-stick, which depends on VLAN tagging? It turns out USB Ethernet adapters generally don’t support VLAN tagging at all. (SNBForums discussion.)

So even if the dongle had been fast and stable — which it was neither — it still couldn’t do the one job I’d bought it for. I’d solved the “second port” problem and walked straight into the “but it can’t do VLANs or stay connected” problem.

There’s also a fourth gremlin worth knowing about: on reboot, USB interfaces can swap names. Your WAN and LAN can quietly trade places, which on a firewall is about as relaxing as it sounds. (OPNsense forum.)

The 11pm reboot of doom

The failure mode that finally broke me was the reboot. I’d make a change, reboot the router, and the USB NIC would either not come back, come back with a different name, or come back just long enough to lull me into a false sense of security before dropping again.

Picture it: it’s late, the house is asleep, the internet is down because I took it down, and I’m crouched over the HP Mini running through the FreeBSD greatest-hits of dmesg errors trying to work out why my second network port has ghosted me. This is the homelab experience they don’t put in the brochure.

What I should have bought

If you’re going to use a USB NIC anyway — and sometimes, in a pinch, you will — at least skip the mistake I made:

Avoid cheap Realtek USB adapters. The community consensus is to prefer ASIX-based adapters (the AX88179 chipset comes up repeatedly), which people report running for months without a single failure. (OPNsense forum.)
Check the chipset and the USB generation, not the price. USB 2.0 can’t carry a gigabit — you want USB 3.0 (5 Gbps) or better as a baseline. The sticker price tells you almost nothing; the chipset and the “3.0” tell you everything.
Treat USB networking as temporary. It’s a bridge to get you running, not a foundation to build on. For anything permanent, you want NICs that live on the actual motherboard or a real expansion card.

But honestly? The bigger lesson was the one I kept refusing to learn: the cheap fix and the right fix are rarely the same thing. I’d now spent money twice — once on a dongle that didn’t work, and once on my own time debugging it — to avoid spending it once on the correct hardware.

This is the point in the story where a sensible person buys a proper dual-NIC router and moves on. Instead, I zoomed out, looked at the whole sprawling mess of router-plus-switch-plus-access-point I was about to assemble from mismatched parts, and thought: what if I just bought a system that was designed to work together?

Enter Omada. Enter, also, a fresh set of problems I didn’t see coming. That’s Part 3.

Previous: ← One NIC and a Dream · Next: Buying the Whole Ecosystem to Solve One Problem →