Posted: 07 Feb 2017 14:32:32
We are seeing issues with IPv6 on a few VDSL cabinets serving our customers. There is no apparent geographical commonality amongst these, as far as we can tell.
Lines pass IPv4 fine, but only intermittently passing IPv6 TCP/UDP for brief amounts of time, usually 4 or so packets, before breaking. Customers have tried BT modem, Asus modem, and our supplied ZyXEL as a modem and router, no difference on any. We also lent them a FireBrick to do some traffic dumps.
Traffic captures at our end and the customer end show that the IPv6 TCP and UDP packets are leaving us but not reaching the customer. ICMP (eg pings) do work.
The first case was reported to us in August 2016, and it has taken a while to get to this point. Until very recently there was only a single reported case. Now that we have four cases we have a bit more information and are able to look at commonalities between them.
Of these circuits, two are serving customers via TalkTalk and two are serving customers via BT backhaul. So this isn't a "carrier network issue", as far as we can make out. The only thing that we can find that is common is that the cabinets are all ECI. (Actually - one of the BT connected customers has migrated to TalkTalk backhaul (still with us, using the same cabinet and phone line etc) and the IPv6 bug has also moved to the new circuit via TalkTalk as the backhaul provider)
We are working with senior TalkTalk engineers to try to perform a traffic capture at the exchange - at the point the traffic leaves TalkTalk equipment and is passed on to Openreach - this will show if the packets are making it that far and will help in pinning down the point at which packets are being lost. Understandably this requires TalkTalk engineers working out of hours to perform this traffic capture and we're currently waiting for when this will happen.
02 Mar 2017 11:14:48
|Packet captures on an affected circuit carried out by TalkTalk have confirmed that this issue most likely lies in the Openreach network.
Circuits that we have been made aware of are being pursued with both BT and TalkTalk for Openreach to make further investigations into the issue.|
If you believe you may be affected please do contact support.
17 Mar 2017 09:44:00
|Having had TalkTalk capture the traffic in the exchange, the next step is to capture traffic at the road-side cabinet. This is being progresses with Openreach and we hope this to happen 'soon'.|
29 Mar 2017 09:52:52
|We've received an update from BT advising that they have been able to replicate the missing IPv6 packets, this is believed to be a bug which they are pursuing with the vendor.|
In the mean time they have also identified a fix which they are working to deploy. We're currently awaiting further details regarding this, and will update this post once further details become known.
18 May 2017 16:30:59
|We've been informed that the fix for this issue is currently being tested with Openreach's supplier, but should be released to them on the 25th May. Once released to Openreach, they will then perform internal testing of this before deploying it to their network. We haven't been provided with any estimation of dates for the final deployment of this fix yet.|
In the interim, we've had all known affected circuits on TalkTalk backhaul have a linecard swap at the cabinet performed as a workaround, which has restored IPv6 on all TT circuits known to be affected by this issue.
BT have come back to us suggesting that they too have a workaround, so we have requested that it is implemented on all known affected BT circuits to restore IPv6 to the customers known to have this issue on BT backhaul.
|Resolution||A fix was rolled out on the last week of June, re-testing with impacted customers has showed that IPv6 is functioning correctly on their lines again after Openreach have applied this fix.|
|Broadband Users Affected||0.05%|
|Started||07 Feb 2017 09:00:00 by AA Staff|
|Closed||19 Jul 2017|
19 Jul 2017
[Broadband] ECI VDSL Cabinets and IPv6 oddness (ICMP OK, TCP or UDP not OK) - Closed