Main branch - server cannot connect to itself

[Nickiel12]

Bug description

I have been having the issues described in issue #1572 , so I switched to the recommended PR (#1564) and I was able to re-register all of my devices, except for the server itself. I then switched to :main to see if that fixed the issue, and after some troubleshooting in the discord, rolled back to the latest stable (0.22.3) which then fixed the issue.

The issue was that when I ran tailscale up --login-server https://<domain> and navigating to the provided URL and running the supplied command, either the node would successfully register, but tailscale would not connect - giving the "go to URL to authenticate" message -, or headscale would fail to add the nodes with the error

github.com/juanfont/headscale/cmd/headscale/cli/utils.go:141 > Could not connect: context deadline exceeded error="context deadline exceeded"

Environment

• Tested using :main 0153e26392506ad389aca2f633e3b638644baa1c and kradalby:1561-online-issue . Issue not preset in stable 0.22.3
• Tailscale 1.54.0, go version: go1.21.4
• Linux, NixOS 24.05 with unstable nixpkgs
• 6.1.62
• I don't know what configuration parameters would be relevant

headscale nodes register output:

2023-12-03T13:30:42-08:00 FTL github.com/juanfont/headscale/cmd/headscale/cli/utils.go:141 > Could not connect: context deadline exceeded error="context deadline exceeded"

I do use nginx to proxy traffic to headscale, however I'm not sure if that counts as a reverse proxy, and this issue was not present when running on the latest stable

To Reproduce

Install headscale on the latest :main version, and try to register the same device running headscale as a node.

[kradalby]

0.23.0-alpha2 addresses a series of issues with node synchronisation, online status and subnet routers, please test this release and report back if the issue still persist.

[Nickiel12]

The host running headscale is able to connect its tailscale client to the server now.