Inaccurate bandwidth reporting for PPPoE on Ubiquiti USG

Tags: #<Tag:0x00007f84b22249b0>

I suspect this is a Ubiquiti issue, rather than LibreNMS, but I’m interested in potential debug steps that might help isolate the cause.

My setup consists of a Ubiquiti UniFi Security Gateway, with LibreNMS performing SNMP polling. Until recently, my WAN interface has been Ethernet from the USG’s eth0 interface to another router, which I believe resulted in accurate bandwidth usage data. I’ve recently had fibre to the premises installed, so the USG’s eth0 interface is now connected to an ONT, but the new ISP requires a PPPoE session to be run over that. I switched the Traffic Bill interface over from eth0 to pppoe0, so as to avoid including the PPPoE encapsulation in the metrics (it’s my understanding the ISP only sees the content, rather than the raw traffic).

The thing is, the gap between reporting on the eth0 interface vs. pppoe0 is huge, which can’t be right. There are no addresses directly assigned to the eth0 interface (except a link-local IPv6, with no peer, or routes), so I can’t see any reason there would be additional traffic beyond the PPPoE tunnel sitting on top of it.

Many thanks in advance.

How much is huge?

Yes, very possible a reporting issue from UBNT.

LibreNMS just stores values given by the device really.

Taking the last 6 hours as an example, eth0 reports a total transfer of 14.82GB vs. pppoe0 showing 159.91MB.

It makes sense that SNMP must be reporting bad value here, but thanks.