Skip to content

fix(platform): the host routes IPv6 through Contabo's router, not fe80::1 - #2639

Closed
ExtraToast wants to merge 1 commit into
mainfrom
fix/2638-host-ipv6-gateway
Closed

ExtraToast wants to merge 1 commit into
mainfrom
fix/2638-host-ipv6-gateway

Conversation

@ExtraToast

Copy link
Copy Markdown
Contributor

Closes #2638

Why

The host's default IPv6 route pointed at fe80::1, where nothing answers. The neighbour table showed fe80::1 FAILED, and the real router answering at fe80::2dc:ff:fe00:2 (MAC 00:dc:00:00:00:02, also 2a02:c207::2). So the host had no IPv6 at all. DNS still publishes its IPv6 address for mail.esa-blueshell.nl, the MX and the web hosts, so senders that prefer IPv6 time out first and mail arrives late.

What changes

The static default route in platform/nix/hosts/frankfurt-contabo-1/default.nix goes via fe80::2dc:ff:fe00:2, still onlink. The comment now names the router address and says why fe80::1 is wrong.

Rollout

  1. Now, without a reboot: sudo ip -6 route replace default via fe80::2dc:ff:fe00:2 dev ens18 onlink. This lasts until the next reboot.
  2. After merge: sudo nixos-rebuild boot --flake 'github:ESA-Blueshell/website/<merge-sha>?dir=platform#frankfurt-contabo-1'. The next reboot keeps the right route. No reboot is needed now, because step 1 already applied it.

Verified

The NixOS flake check CI job builds the config. The route itself is verified on the host by step 1 and a ping -6 to 2606:4700:4700::1111.

…0::1 (#2638)

Nothing answers on fe80::1, so the default IPv6 route led nowhere: the host
had no IPv6, while DNS publishes its address for mail and the web hosts. The
neighbour table shows the router at fe80::2dc:ff:fe00:2.
@github-actions

github-actions Bot commented Oct 27, 2026 •

Copy link
Copy Markdown
Contributor

Changes

Bucket Files Added Removed
platform 1 █████ 7 ░░░░░ 7
total 1 +7 −7

No production or test lines added.

Coverage

No coverage summary reached this report.

Patch coverage: this pull request changes no line that coverage measures.

No coverage reached this report for api (api-unit-test-reports).
No coverage reached this report for frontend (frontend-unit-coverage).
No coverage reached this report for imaging (imaging-unit-coverage).

@ExtraToast

Copy link
Copy Markdown
Contributor Author

Superseded: the gateway was not the cause (both routers resolve, nothing is routed), and the cluster is IPv4-only anyway. IPv6 is switched off in #2641 and done properly in #2640.

@ExtraToast ExtraToast closed this Oct 27, 2026
@ExtraToast
ExtraToast deleted the fix/2638-host-ipv6-gateway branch October 27, 2026 14:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Platform: stop advertising IPv6 until the server can serve it

1 participant