Andrew Zonenberg on Nostr: Whyyyyy is the world of IPv6 (on Comcast at least) allergic to static configuration?? ...
Whyyyyy is the world of IPv6 (on Comcast at least) allergic to static configuration??
I have a static /56 allocation so this SHOULD be straightforward:
1) Hand out /64's from this allocation to each local subnet
2) Set router's default route to the v6 address of the Comcast CPE
3) Add a static route on the CPE to send all traffic to my /56 to my router
But no.
1) You *have* to do DHCPv6 PD. If you don't ask the CPE for a delegated prefix, there's no way to tell it to route that traffic to you.
2) Something seems buggy around PD lease renewal, maybe the actual enforced expiry timer doesn't match what it claims in the advertisement? The end result is that the CPE randomly stops passing IPv6 traffic until you restart the DHCPv6 client service.
3) The CPE is using IPv6 privacy addresses or similar on its local interface (or randomly rotating MACs or something). This makes static next-hop routing impossible, as a few hours after you get your config working your default gateway won't be on that IP anymore.
The only workaround I've found is to force my router's accept_ra sysctl on the upstream interface to 2, allowing it to accept IPv6 RAs even when it's configured with ip_forward.
Why is a *router* randomly changing IP addresses on me? This stupidity should be illegal. If you want to track me, the /56 is static and unique to the street address anyway. There is absolutely no point in changing the low 64 of my next-hop route.
I have a static /56 allocation so this SHOULD be straightforward:
1) Hand out /64's from this allocation to each local subnet
2) Set router's default route to the v6 address of the Comcast CPE
3) Add a static route on the CPE to send all traffic to my /56 to my router
But no.
1) You *have* to do DHCPv6 PD. If you don't ask the CPE for a delegated prefix, there's no way to tell it to route that traffic to you.
2) Something seems buggy around PD lease renewal, maybe the actual enforced expiry timer doesn't match what it claims in the advertisement? The end result is that the CPE randomly stops passing IPv6 traffic until you restart the DHCPv6 client service.
3) The CPE is using IPv6 privacy addresses or similar on its local interface (or randomly rotating MACs or something). This makes static next-hop routing impossible, as a few hours after you get your config working your default gateway won't be on that IP anymore.
The only workaround I've found is to force my router's accept_ra sysctl on the upstream interface to 2, allowing it to accept IPv6 RAs even when it's configured with ip_forward.
Why is a *router* randomly changing IP addresses on me? This stupidity should be illegal. If you want to track me, the /56 is static and unique to the street address anyway. There is absolutely no point in changing the low 64 of my next-hop route.