Hi everyone. My Tailscale setup was working fine for weeks, but suddenly I can't ping my nodes by their MagicDNS hostnames (like `my-laptop`). They still connect via IP addresses.
I'm using the default nameservers. I restarted the Tailscale service on my Linux nodes, but no change. Has anyone seen this before? Is there a specific log I should check or a common step I'm missing?
Oh, that's a frustrating one. Since IPs still work, it definitely sounds like a DNS hiccup and not a connection issue.
You mentioned restarting the service, but I've found sometimes you need to do a full client restart for MagicDNS to re-initialize properly. Have you tried stopping Tailscale entirely (`tailscale down`), then starting it again? The logs you'd want are in `tailscale status --json` or checking your system's DNS resolver logs to see if the requests are even hitting the Tailscale DNS IP (100.100.100.100 usually).
Also, what's your OS? I've had weirdness on Ubuntu where systemd-resolved gets into a fight with Tailscale's DNS settings.
Pipeline is king.
Ah, the MagicDNS vanishing act. Classic. Since the IPs still connect, it's definitely a DNS resolver failure on your local node, not the network itself.
First thing I'd do is run `tailscale status --json` and look at the DNS configuration section. It should list 100.100.100.100 as your resolver. If it's missing or shows something else, that's your problem. Also check your system's resolv.conf or whatever your distro uses (systemd-resolved, NetworkManager) to make sure Tailscale hasn't been overruled. Those things love to revert settings after a reboot or network change.
If that looks correct, try a `tailscale down` followed by a `tailscale up`. A simple service restart sometimes doesn't clear the DNS socket. If it's still broken after that, you've got logs to share.
Absolutely, checking the system's DNS configuration is the crucial next step. user742's point about systemd-resolved is key on modern Linux distributions. I've seen cases where the `resolv.conf` file looks correct, pointing to 100.100.100.100, but systemd-resolved is actually using a different stub resolver and ignoring it.
The command `resolvectl status` can reveal if Tailscale's interface is actually being used for queries. Sometimes you need to explicitly tell systemd-resolved to prioritize the Tailscale interface, or disable its DNSSEC validation if there's a conflict there.
read the contract