Signup fails, sign in succeeds with error. Is your IPv6 broken?

[SleeplessByte]

ℹ️ I do not need help signin in or up, as flyctl auth whoami returns my e-mail address. The sign up experience was jarring, and there seems to be some sort of issue with connecting to api.fly.io on Windows. This may definitely be a problem with my setup, but given that this is an almost brand new installation of the OS without any extra firewall setup, this may be a problem for more people.

Microsoft Windows 11 Pro
Version 10.0.22000 Build 22000

No extra firewall
No pihole
No custom DNS

Describe the bug

Following https://fly.io/docs/getting-started/log-in-to-fly/ I am on Windows 11.

Installing

iwr https://fly.io/install.ps1 -useb | iex in PowerShell. Succeeds without warning

flyctl was installed successfully to C:\Users\Derk-Jan.fly\bin\flyctl.exe
Run 'flyctl --help' to get started

Sign-up

So I wanted to sign up:

$ flyctl auth signup
Error Post "https://api.fly.io/api/v1/cli_sessions": read tcp [<snip>]:55807->[<snip>]:443: wsarecv: An existing connection was forcibly closed by the remote host.

I tried a few more times, also in a new shell. Could not get it to work.

I tried flyctl --help instead, to see if it was correctly installed (it was), and then ran flyctl auth --help to find out I could just try to login

Login

$ flyctl auth login
WARN failed querying for new release: Get "https://api.fly.io/app/flyctl_releases/windows/amd64/latest": read tcp [<snip>]:54496->[<snip>]:443: wsarecv: An existing connection was forcibly closed by the remote host.
Opening https://fly.io/app/auth/cli/<snip> ...

I was shown an error (similar to the error above) but the browser automatically reloaded (instantly) and showed me a sign in page. I pressed Sign In with GitHub (which I had never done before), and authorized fly.io.

I was redirected to an "empty" page with an error message:

CLI Session Expired, please restart the process

However, on my PowerShell:

Waiting for session... Done
successfully logged in as derk-jan<snip>@<snip>

Retrying the command to login shows the same warning for "failed quering for new release" but allows me to login again with my GitHub (via a quick button) and then it goes throw without CLI Session Expired error.

Trying DNS settings to resolve

When configuring ipv4 DNS by CloudFlare (which should synthetically resolve IPv6 as well), https://test-ipv6.com/ gives a 10/10 and https://ipv4.google.com/ loads correctly meaning both IPv6 and IPv4 are correctly working. However, if I now try to whoami:

failed retrieving current user: Post "https://api.fly.io/graphql": read tcp []:55057->[2a09:8280:1:f28:246e:d6a:949:dbbf]:443: wsarecv: An existing connection was forcibly closed by the remote host.

There seems to be something wrong with the IPv6 connection on fly.io's end, which seemed to happen in Chrome too right before it reloaded. I have no clue how to further debug or resolve this, and I am fine relying on manual updates and turning off CloudFlare DNS for now, but this seems like something that's not actually broken on my end.

Let me know if I can provide further information to revolve this.

[jsierles]

Thanks for the report. This repo is for more specific flyctl bugs, and is not monitored by that many people. Can you repost this on https://community.fly.io? Thanks!

[SleeplessByte]

Will do!

[michaelbromley]

Dropping this here as we ran into the same issue, and there was no solution to be found on the community forum:

The solution for us was found in the following tweet from a Fly.io employee: https://twitter.com/mrkurt/status/1533489925575872512

Maybe try ipv4? You can make a hosts entry pointing http://api.fly.io -> 77.83.143.220 to do that, I think.

[SleeplessByte]

Note that that isn't a solution but a workaround and will break if api.fly.io every moves.

I gave up on fly.io not long after this.

[jonhoo]

For anyone else who ends up here, this appears to be because fly and flyctl both require a DNS resolver to be running [::1]:53 (i.e., on the IPv6 localhost address). To force systemd-resolved to do that, you can add

DNSStubListenerExtra=[::1]:53

to /etc/systemd/resolved.conf (see this SO answer).