Debian/Ubuntu not keeping tags and routes after reboot

I haven’t found a clear link as to when or why it happens, but this issue doesn’t happen consistently for me.

I have a number of Debian and Ubuntu machines connected with various custom parameters.
For example, my DNS server (PiHole) is tagged with tag:dns and doesn’t accept routes or dns from other machines.
My subnet routing machine has a bunch of routes associted with it.
Some VPS machines are just tagged and set to not accept routes/dns.

The problem is that when I reboot, sometime I lose these tags, routes, etc. The machine with subnet routes will show up like a change and say that its no longer advertising them. The machines with tags will just drop the tag. etc.

I’ve tried re-doing the “tailscale up” command for that machine (with routes, tags, etc.), which usually does nothing to add tags, routes, etc. So I have to append --reset to get it to register properly. Then I reboot again and it goes away, sometimes.

Some example command lines I have used on these machines (I only used the --reset flag when I couldnt get it to register properly).

sudo tailscale up --accept-routes=false --accept-dns=false --reset

sudo tailscale up --advertise-routes=192.x.x.0/24 --advertise-exit-node --accept-routes=false --accept-dns=false --reset

Any ideas why I keep losing network functionality for rebooting a machine?

I do have some windows machines, but only one has a tag associated with and I would need to test it further, so I can’t comment on those.

Something I noticed that has worked so far has been to go on each machine with custom routes, tags, etc. and issue a “sudo tailscale down” command. Then re-issue to associated “sudo tailscale up” command without --reset in it. So far, each machine I’ve tried this on has survived a reboot. I can’t prove this, but it seems like the service was picking up old command line flags when it restarted so maybe I had to “fix” them by tailscale down and then up?

Just curious: When you are having issues with settings persisting, does Tailscale come up, just without the settings, or does Tailscale not come up at all after reboot?

I’ve seen both. On my subnet routing node, it would restart, show connected on the dashboard, but show a little circle with a dash in it on the routes. When I review the routes, it says they aren’t being advertised anymore and I have the option to delete/disable them. A tailscale up with the proper flags was required to get it working again.
Other nodes would restart and come online without tags associated anymore, but connected normally. Then a few times, the device did not come back online at all after a reboot (Tailscale didnt come up, but the device was online normally), but just required a tailscale up --flags command to fix it.

Today, I can’t reproduce any of it. The biggest change was doing a --force-reauth with the proper flags the last time I re-enabled it.

I have an open GitHub issue on this as well.

Thank you for this tip. I had an issue where Tailscale simply would not autostart on reboot.

The --force-reauth seems to have done the trick on both of my RaspberryPis.

Thanks again!

I had a similar issue that after doing a apt-get dist-upgrade and upgrading tailscale by a minor version it wouldn’t come up on its own - if you guys update and then run tailscale status will it report as up and running?

I just did apt update followed by apt upgrade and went from Tailscale 1.10.1 to 1.10.2 on Raspbian Buster on 2 Pis, and after manual reboot, tailscale status still shows as running.

Just curious if you tried the --force-reauth and then reboot ?

This was fixed in the just-released 1.10.2. There many dups of this bug but the one where the clues came together for us was in #2321.

Before upgrading to 1.10.2, if you run tailscale down; tailscale up it’ll force saving the prefs to disk so they won’t be lost on upgrade.

ah, thanks. I will check again when the next update is out :slight_smile: