Inbound traffic suddenly stops

[#1]

---

@DavidJ wrote:

Yes, just the single cable between the UDM and the Vigor's P1 port. Nothing plugged into the P2 port.

Thanks.

So you are connected the same way as you would have had your Vigor 130 to access the GUI,  because the 130 only has one ethernet port.

In the light of @MisterW 's theory, I wondered whether it is worth trying the Vigor 166 with separate paths for PPPoE and the GUI connections, which would make packet capture cleaner to view, and it 'might' fix whatever the problem is.

So you'd have to change the Vigor 166's LAN IP address to a value within your router's LAN subnet

Then connect the Vigor 166's ethernet port "P2" to an ethernet switch outlet on your router's LAN network

The Vigor 166's port "P1" should still be connected to your UDM's WAN port (for PPPoE)

For completeness, you could remove the static IP address from the UDM's WAN port, and delete the associated static route

Now your UDM WAN port will only pass PPPoE traffic, and the Vigor 166's GUI can be accessed from your LAN, with the PPPoE and GUI traffic on separate ethernet ports.

[#2]

---

@DavidJ wrote:

There are a couple of things I'd like to try with the Vigor before I swap it out again. It gives the options of 2 different "dsl modem codes" which I believe are dsl modem firmware versions. I'll try the other version to see if that makes any difference.

When you get around to testing the different modem codes, also double check that -

        "DSL Mode:" = "VDSL2 only"

        "Hardware Acceleration" = "Disable"

[#3]

Some more testing has highlighted the issue, if not exactly why it's happening...

I swapped the Vigor 166 back and tried the different DSL modem codes but that made no difference.

So I factory reset my Vigor and made only the minimum essential configuration. Result - everything works fine. It has been suggested that the problem might be related to the routing configuration that was used to permit access to the Vigor's GUI from my lan. So I reinstated that configuration. Result - everything still works fine.

So I ran through the rest of the Vigor's configuration and the only other setting that was different was Hardware Acceleration. I've always had that set on but the factory reset had turned it off. I turned it on. Result - the problem returned - regular disconnects/reconnects. I turned it off and the problem went away.

So there we go - the issue seems to be the hardware acceleration setting on the Vigor. As I said, I've always had it on and there's never been a problem so I don't know what's changed. And how the different firmware versions on the UDM trigger the issue is still a mystery.

@Anonymous , I see you suggested the same thing. Great minds obviously think alike!

@MisterW , I did a packet capture when I reinstated the routing configuration to access the Vigor's GUI. You were absolutely right - all internet bound traffic is sent with a destination mac address of something beyond the Vigor (the 'concentrator' I believe). Traffic generated to access the Vigor's GUI has the destination mac address of the Vigor's LAN port.

I will raise this issue with Draytek and report back what they say.

[#4]

@DavidJ 

Excellent diagnostics, though if this had been posted yesterday, I'd be suspecting some leg pulling...

• Some event in the router upsets the modem
• Old router firmware not too much of a problem
• Rebooting the router (leaving sulking modem as is) resets the problem until next time
• New router firmware a substantial problem all of the time
• Turn off hardware acceleration on the modem and the problem (apparently caused by the router) vanishes entirely

That is some wicked interaction!  Will be very interesting to see what answer comes from the vendor(s) if any.

[#5]

I've raised it with Draytek and will report back.

[#6]

So there we go - the issue seems to be the hardware acceleration setting on the Vigor. As I said, I've always had it on and there's never been a problem so I don't know what's changed. And how the different firmware versions on the UDM trigger the issue is still a mystery.

@DavidJ Hardware acceration is usually associated with NAT performance. The h/w maintains tables of NAT connections and allows incoming packets to bypass much of the software , thus enabling higher performance. My Openwrt router has the same facility and allows around 940Mb/s NAT throughput with it enabled whereas it can only achieve approx 400Mb/s with it disabled.

The Vigor 166 can be a router and so one can see why this option would be there. However in modem mode, I cant see what it would do, there is no routing and by implication no NAT!. Why it is even an option in modem mode I cant really understand ?

https://www.draytek.co.uk/pdf/10722_Draytek_datasheet_Vigor166.pdf

It'll be interesting to see what Draytek say...

[#7]

The Vigor 166 can be a router and so one can see why this option would be there. However in modem mode, I cant see what it would do, there is no NAT!. Why it is even an option in modem mode I cant really understand ?

@MisterW 

I had wondered what benefit it had in bridge mode. However, I assumed that because the option was still available when in bridge mode, there must be some benefit. Having said that, my VDSL line gives me about 70mbps so I don't think hardware acceleration, of the lack of it, is an issue.

[#8]

@DavidJ yes its almost ceratinly no loss on your connection. The specs claim that it can do 700Mb without acceleration so unless its being used on a VERY high speed G.Fast connection (in router mode) then acceleration is of no use.

[#9]

@MisterW , @Townman , @Anonymous 

An update...

Heard back from Draytek. Their initial response was "are you sure you're in modem/bridge mode?", which suggests hardware acceleration really isn't doing anything in bridge mode and possibly/probably shouldn't be an option in that mode. Having assured them I was indeed in bridge mode, their reply was "I'm off to the lab" but I haven't heard anything back.

Heard back from Ubiquiti. They seem a little more interested and have asked me to do more packet captures when the UDM is triggering the problem (i.e. with hardware acceleration enabled) and when it isn't (i.e. with hardware acceleration disabled). I'm not sure how fruitful this is going to be as it's possible the problem has something to do with the Vigor's disliking of my line (remember, my old Vigor 130 never got on with my line and was swapped for the 166 by Draytek). It may only be something that could be properly investigated by Ubiquiti and Draytek together and, given that there's a solution, I can't imagine this is something on which they're going to spend a lot of time. Plus, I'm sick of the site of Wireshark!

Also, following a comment by @Anonymous I got hold of a Billion 8800NL R2 router from eBay. It supports bridge mode and has a Broadcom chipset as opposed to the Lantiq which I believe are contained in Drayteks. Looking at the information reported by my Vigor 166, I had noticed that my line is connected to Broadcom hardware at 'the other end'. Having seen various reports of certain modems/routers working best with certain xDSL chipset manufacturers, I was interested to see what difference this might make. Not only that, but the Billion router gives more stats on WAN traffic which, if that had been available from the Vigor, might have made diagnosing the original problem easier.

Anyway, I disabled wireless, whipped off the aerials and stuck it into bridge mode (i.e. turned it into the equivalent of the 166). Everything works. Moreover, I noticed that my line sync'd at 79999/19999 (it's an 80/20 line). The best I ever got with the Vigor was 76-ish on the download (I always got 19999 upload). It doesn't really make a noticeable difference on a speed test but I'll take 79 over 76 if it's available. Would this be down to Broadcom chipset talking to Broadcom chipset or are there other factors? No big deal, just interested.

If/when I hear more from Draytek and Ubiquiti I'll report back.

And once again, thanks for all your help with this. It's been a weird one! I think @Townman said it best: "That is some wicked interaction!".

[#10]

Their initial response was "are you sure you're in modem/bridge mode?", which suggests hardware acceleration really isn't doing anything in bridge mode and possibly/probably shouldn't be an option in that mode. Having assured them I was indeed in bridge mode, their reply was "I'm off to the lab" but I haven't heard anything back.

😀

Sounds about right, hardware acceration is all about offloading the NAT translation from the processor to hardware.

Since in modem mode there IS no NAT, I'm not surprised they've gone into a huddle...

[#11]

Of all the FTTC modems and modem//routers I've tested, the 8800NL-R2 gives the most stable and fastest sync speeds on it's default configuration.  However using the SNR target tweak in a Vigor 130 can beat the 8800NL, but requires some knowledge to get it right.

I think that Broadcom based modems simply work better than Lantiq, but matching the chipsets of modem and cabinet is also likely to help.

Have you put the last known firmware in to the 8800 ? - 2.52.d18  

I'm glad that the 8800NL-R2 worked out well for you !

I would be using that if the PPPoE connection supported Jumbo Frames (giving a WAN MTU=1500 rather than 1492)

I did put in a feature request to Billion for a firmware update with Jumbo Frames, but they've not released any updates since my request.

.

[#12]

@Anonymous 

Have you put the last known firmware in to the 8800 ? - 2.52.d18  

---

First thing I did.

Thanks for all your help.