BQM ping dropped, everything else can ping my IP fine

I've noticed this with PPPoE and L2TP recently, after a restart of my router my BQM for a static IP on a VDSL line is currently showing 100% loss, but the IP can be pinged fine from anywhere else. I can even see in real-time the BQM source IP pinging my static IP.

80.249.99.164 87.xx.xx.xx 117.80 KB (4308 Pkts.)

There is no rate limiting being applied, this is an OpenWrt router, so I have full control on what it happening on the firewall. This doesn't happen with my Virgin Media IP on the same router, which is being pinged by the same host as we speak.

If I delete the BQM and recreate it with the same IP it will work, as I've tested that before, but obviously that will lose historical data. I've even tried reloading my firewall, but as everything else can ping, it doesn't seem like it would do much.

I don't understand though why the BQM pings would timeout when everything can ping fine.

Edited by jamesmacwhite (Sun 20-Sep-20 21:25:01)

I've figured out what's happening, due to being multi homed with multiple WAN interfaces, it would appear the echo reply response to the ping responses from pingbox1.thinkbroadband.com are going over the wrong WAN interface, so it's a routing problem.

Analysing with tcpdump, the echo reply is going over another WAN interface.

Edited by jamesmacwhite (Sun 27-Sep-20 13:36:14)

In reply to a post by jamesmacwhite:

Analysing with tcpdump, the echo reply is going over another WAN interface.

Asymmetric routing is sometimes useful, but in this case you need the reply to go back out of the interface the request came in from.

In reply to a post by jchamier:

In reply to a post by jamesmacwhite:

Analysing with tcpdump, the echo reply is going over another WAN interface.

Asymmetric routing is sometimes useful, but in this case you need the reply to go back out of the interface the request came in from.

Yes, it makes sense now after seeing the ICMP traffic with tcpdump. For some reason though, it was specifically ICMP requests from TBB that was going over the wrong interface, everything other ICMP test I did was fine.

I use OpenWrt with mwan3, so it's possible it's an issue to do with how it handles routing. The echo reply does indeed need to go over the same interface. The replies were happening just not over the right interface.

I will have to see if I can make sure ICMP traffic specifically stays on the specific interface.