Tailscale DNS settings are overwritten every now and then

[guilhermearaujo]

I used the Tailscale script to add severa LXCs to my Tailnet, and off the bat everthing seemed fine. The machines were registered, and one container could connect to anothers' services.

I am using Debian 12 (bookworm) in all containers

However after some hours, usually, the file /etc/resolv.conf is replaced by dhclient with the default nameserver provided by my router, instead of Tailscale's 100.100.100.100. From this event on, this container will not be able to resolve "machine-ids" to their IPs as they did before. I need to run tailscale down; tailscale up for the DNS settings be updated back to Tailscale's.

I found this documentation page acknowledging this as a DNS fight that occurs particularly with dhclient, but not with other services such as systemd-resolved.

I found instructions to install systemd-resolved, but I am not sure how to make sure dhclient is/will be disabled.

Also, if this is really that simple, perhaps the script could detect which DNS method is being used and suggest & perform the replacement. Do you think it would be possible?

[guilhermearaujo]

Installing systemd-resolved didn't fix. They are still fighting. For now, I added a cronjob to restart tailscale every hour, but it's far from ideal.