Appearance
Tailscale
Tailscale provides remote access through a subnet router and optional internet exit node. It runs natively in the VPN LXC; the LXC's TUN device and forwarding configuration are prerequisites.
Private identity handling, account details, and AI operations are in Tailscale · Internal Operations.
Access and configuration
| Setting | This homelab |
|---|---|
| Guest | LXC 102, vpn |
| LAN address | 10.120.5.9/24, VLAN 5 |
| Gateway | 10.120.5.1 |
| Native service | tailscaled.service |
| Tunnel interface | tailscale0 |
| Subnet routes | 10.120.5.0/24, 10.120.10.0/24, 10.120.30.0/24, 10.120.40.0/24 |
| Exit node | Advertised, separately approved and selected by clients |
There is no public web dashboard or reverse-proxy URL for this guest. Use your tailnet's admin console and local CLI. For another lab, advertise only your intended networks and confirm the router can reach them. The lab's guest ID does not match its address's last octet.
Install and enroll
Prepare the VPN guest, then install from the official Linux download instructions. The lab uses the stable Ubuntu Noble APT repository. You can download and inspect the official installer before running it:
sh
curl -fsSL https://tailscale.com/install.sh -o /tmp/tailscale-install.sh
less /tmp/tailscale-install.sh
sudo sh /tmp/tailscale-install.sh
sudo systemctl enable --now tailscaled
sudo tailscale upFollow the generated login URL in a trusted browser and enroll the node in your intended tailnet. For unattended enrollment, use a short-lived, one-off auth key with the appropriate tag and approval policy; keep it out of shell history and documentation.
A restored complete LXC backup already contains a node identity. Do not enroll it as another device unless that identity is intentionally being replaced, and never run two active copies of the same restored identity.
Enable IP forwarding
Create /etc/sysctl.d/99-tailscale.conf:
text
net.ipv4.ip_forward = 1
net.ipv6.conf.all.forwarding = 1sh
sudo sysctl --system
sysctl net.ipv4.ip_forward net.ipv6.conf.all.forwardingBoth values should be 1. The host must route traffic beyond the tunnel as described in the Tailscale forwarding documentation.
Advertise routes and the exit node
Keep a direct LAN management connection or Proxmox console available before changing routing. This example reproduces the lab's routing intent:
sh
sudo tailscale set \
--accept-routes=false \
--advertise-routes=10.120.5.0/24,10.120.10.0/24,10.120.30.0/24,10.120.40.0/24 \
--advertise-exit-node=true \
--snat-subnet-routes=true \
--netfilter-mode=onReplace the route list for your own network. Each --advertise-routes value replaces the complete list; omitted subnets are withdrawn. Internet default routes are handled by the exit-node option rather than added to this list. SNAT means internal services see the subnet router's source address instead of the remote client's tailnet address.
In the Tailscale admin console, select this device and approve the intended subnet routes and exit-node use. Separately configure tailnet access policy for the users and destinations that should be reachable. Route approval does not grant access by itself. Ensure your local gateway routes between the VPN guest's VLAN and the target VLANs.
Linux clients needing subnet routes must enable acceptance with sudo tailscale set --accept-routes=true. Select the exit node on clients only when you want their internet traffic routed through the lab.
Optional preferences
The lab accepts Tailscale DNS, enables Tailscale SSH, disables the web client, and enables automatic update installation. Configure these features deliberately after enrollment. Tailscale SSH also requires a matching tailnet SSH policy; enabling it does not configure that policy. Keep ordinary key-based LAN SSH as a separate recovery path.
Use tailscale set for individual preference changes instead of rerunning an enrollment command with a new auth key. Keep the exact host-specific preference set and identity recovery in the internal runbook.
Verify routing
On the VPN guest:
sh
systemctl is-active tailscaled
tailscale status
tailscale netcheck
ip -brief address show eth0 tailscale0
sysctl net.ipv4.ip_forward net.ipv6.conf.all.forwardingFrom a separate remote tailnet client, ping the node and test a permitted service in every advertised subnet. Select the exit node temporarily, verify internet access and the expected egress address, then restore the client's prior choice. Local tailscale status alone cannot prove remote forwarding works.
Updates and recovery
Restarting tailscaled interrupts subnet routes, exit-node traffic, and Tailscale SSH through this node. Use a direct management path during maintenance. Update packages through the configured Tailscale repository or deliberate auto-update policy, and repeat remote checks afterward.
Back up the LXC off-host before risky routing changes. /var/lib/tailscale/tailscaled.state contains its private identity: protect the backup and never print that file. Restore a known-good guest with the intended network configuration, then verify TUN access, forwarding, route approvals, access policy, and local gateway reachability before considering re-enrollment.