IPv6 Trial Update

[#1]

All switched on for me last night. Working great, 10/10 on IPv6 test website. I have a Unifi USG-Lite. Had to enable IPv6 on the WAN (set to DHCPv6, entered Prefix Delegation as "56") and then enabled on the LAN network(s) too.

When you get around to a public release and put out generic instructions for third party routers, it would be helpful to state that DHCPv6, PD and 56 needs to be entered, as I naively tried SLAAC first which was incorrect.

[#2]

entered Prefix Delegation as "56"

@AndrewN is there not a setting for 'automatic' ? so the prefix length is defined by the ISP

[#3]

---

@MisterW wrote:

entered Prefix Delegation as "56"

@AndrewN is there not a setting for 'automatic' ? so the prefix length is defined by the ISP

---

I was thinking the very same thing but looking at this it sounds like you do indeed need to know and set it. Seems an odd thing to require given that it isn't a mandated requirement of the underlying protocol. Indeed the client can only hint at what prefix length it would like - the choice is entirely down to the server. Might be worth @AndrewN  leaving it blank and seeing if the router accepts this and can still obtain a prefix.

[#4]

It’s a required field and cannot be left blank.

[#5]

Ah okay.

[#6]

I did post this earlier in the thread, as a hopeful hint for other fellow Ubiquiti users

Edit: I also enable IPv6 on the LAN side with Prefix Delegation (set to Auto) and SLAAC; my brief research summarised that internally SLAAC is preferred over DHCPv6 for home use.

[#7]

I have two options for SLAAC on the LAN side:

SLAAC + stateless DHCP and 

SLAAC + RDNSS

Both seem to work but I'm unclear about the difference and which, if either, is preferable. Any pointers?

[#8]

'It depends' on what devices you have active on your home network.

Wikipedia has a bit of a list about what devices support RDNSS:

> https://en.wikipedia.org/wiki/Comparison_of_IPv6_support_in_operating_systems

Stateless DHCP is thought to be more compatible with older devices.

I guess ultimately if you find a device that doesn't appear to be working correctly, try changing to the other way of providing DNS data to the network.

Seems a bit odd that your router can't enable both, as I don't think they incompatible with each other if both are active at the same time (and providing the same DNS data).

https://tolumichael.com/slaacstateless-dhcp-vs-slaacrdnss/ has a deeper dive about it.

[#9]

@IMM I'd suggest SLAAC + RDNSS for home use.

From my earlier research, one of the smaller reasons was that Android devices and Chrome OS don't support DHCPv6. Google refuse to support it for some reason.

[#10]

I am actually having issues here now. I get 10/10 on https://test-ipv6.com/ and yet loading web pages on my Windows devices is sometimes slow. Bandwidth is fine (900/220) but browsing my mailbox in Outlook Web Access and various other websites leads to very slow loading times for elements on the page - images, email content etc. and sometimes just straight up "network reset" when attempting to access a website. If I disable IPv6 on my Windows PC, it loads just fine. It seems to be only a Windows issue - fine on the same websites on my iPhone (it is definitely using IPv6).

Has anyone experienced similar issues after enabling IPv6?

I'm using a Ubiquiti Unifi USG-Lite and have DNS set to Cloudflare & Cloudflare-ipv6 (also tried Google variants, but same behaviour).

I doubt this is Plusnet specific issue, so I'll have a hunt around the web!

[#11]

Hi! If you're still taking people on for the trial, I'd be interested in taking part! I have a static IP, Plusnet Hub 2, and running on Linux Mint. Thank you:)

[#12]

When the trail completes and (if) IPv6 is rolled out, will Plusnet offer a static IPv6 allocation to customers?

[#13]

@AndrewN not heard anything similar but I don't use Windows at home, will try and big out my old Macbook from the garage and see if still have bootcamp running and see if I can see anything.

@Wonder I'll let you know if we are able to add anyone else.

@rflitcroft we've not made any decisions yet, the original plan was to have both a dynamic (more sticky than dynamic because you'll likely to keep getting the same prefix but it's not guaranteed to stay the same) and static IP trial running together, dynamic is a bit behind schedule at the moment unfortunately. There's also a number of pieces of backend development we need to do to support the rollout so the end of the trial is still a while off and hopefully I'll be able to give you a better answer in a few months.

[#14]

@dave great, thank you!

[#15]

@AndrewN 

Having run ip6 via a HE tunnelbroker for years, I did see some of the 'random slowness compared to IP4 NAT' behaviour.

Sometimes there are MTU issues with particular sites - a lack of understanding in some site operators about allowing (necessary) ICMPv6 means that path MTU discovery was/is sometimes an issue.  When it was really bad, I dropped the IP6 MTU to 1280 (minimum that's required by required to be supported by IP6) to side-step broken sites.  I'm back to 1460 now as that seems to generally fine now.  Cloudflare have some articles about this and what they do to mitigate issues.

The HE tunnels were increasingly bad as being flagged for the 'Are you human' checks by the cloudflare systems that front up many web sites.  It's not supposed to happen for non-interactive services, but that can also cause issues and stalls until the browser gives up and retries.

There's another issue I was seeing with Linux 6in4 tunnels that would drop random packets an impact throughput (ECN issues where the IP6 packet header would be mis-interpreted as IP4 ECN and cause drops), but that's not going to be an issue with Windows + PlusNET's native IP6.

The path MTU discovery issue might be what's hitting you.  Drop the MTU to 1280 and see if that goes away.  ( There's a few percent hit from the lower data-to-header ratio, but you may see the stalls go away. )

Good luck!

Mark