After booting Windows 10, Tailscale connections fail until I reconnect Tailscale

Whenever I boot Windows for the first time, my machine can’t connect to any other Tailscale IPs. And other machines can’t connect to me. It’s as if Tailscale is down. But tailscale ping some-other-machine works correctly.

But I found a workaround: right-click on the Tailscale icon in the bottom-right taskbar, click “Disconnect”, then “Connect”. Incoming and outgoing Tailscale connections work perfectly after that.

I originally thought this was a Windows Firewall issue, but the disconnect/connect workaround confuses me. How do I continue troubleshooting? Is this a known issue? I’d like to stop using this workaround because it won’t work on unattended boots, and it’s annoying to remember after a reboot.

I’m running Tailscale 1.26.0 on Windows 10 Pro (10.0.19043 Build 19043). I also have some network adapters related to TAP, VPNs, and VirtualBox installed.

This problem doesn’t happen on another Windows 10 machine (a laptop) that I also use.

This still happens to me on Tailscale 1.28.0 on every boot.

Also getting the same issue and using the same workaround, running Tailscale 1.28.0 on Windows 10 Enterprise (Build 19044.1889).

Without the workaround, neither the Tailscale IP nor the MagicDNS name works for addressing, but using tailscale ping machine-name returns a pong as intended.

Tried doing:

  • Setting the Tailscale service to “Automatic (Delayed Start)”
    • The GUI launches on startup, but shows an error popup that the service failed
  • Enable “Allow service to interact with desktop” in the service
    • No change
  • Add option to restart service after 1 minute of failure
    • No change (the service isn’t failing, it just couldn’t connect to the machine)

My guess is that the service starts successfully, but nonetheless fails to load Tailscale’s routes into Windows on startup; maybe it’s trying to establish the routes while the network isn’t ready yet (drivers still loading, network still connecting, etc.)?

I’m also having the same issues on the latest version when resuming from sleep.

After upgrading to Windows 10 21H2, I no longer experience this issue.

Still experiencing this issue, using Tailscale 1.30.2 on Windows 11 Enterprise 21H2 (Build 22000.1042).


Ping results using PowerShell before disconnecting and reconnecting:

image

That’s interesting, since this fixed itself for me somehow.

I’m on Windows 10 Pro Build 19044.2006. But I had updates enabled for the last few days, and I’m not sure what build number I had last week.

Found the source of the issue using Tailscale 1.30.2 by running tailscale status: it fails to append to the etc\hosts file on launch with an Access is denied error.

Running tailscale down then tailscale up or restarting with the tray icon some time after startup makes the host append work.

This happened to me even on Tailscale 1.34.1 (Windows 11 22H2 22621.963) but anyway.
What works though it’s not a workaround but I’m willing to live with it is:

After reboot, my user logs in, tailscale service starts and I can ssh into the machine.
Without needing to Disconnect and (Re)Connect.

Works for me for the moment but it sounds like something Tailscale should fix on Windows about Run Unattended which is what they declare to support Keep Tailscale running when I'm not logged in to my computer · Tailscale

I’m on Tailscale 1.38.4 on Windows 10 LTSC 2021.

Without automatic logon, Tailscale doesn’t connect to the tailnet on boot. Even with “Run Unattended” enabled.

I followed @mnz 's suggestion to configure automatic login. Now it works, but I noted that:

  • I didn’t have to disable “Run Unattended”
  • I also didn’t have to modify the Tailscale service settings to “Allow service to interact with desktop”

Maybe something has changed in the meantime (1.34 → 1.38), but still – many thanks to @mnz !

And yeah, this should be fixed so that automatic logon does not need to be enabled.