No DHCP on same subnet with Tailscale installed

[jelmervdm]

Hi,

I have 6 DHCP scopes configured, one for each VLAN. DHCP is routed through my switch with DHCP relay enabled. DHCP for my server network is working, I only use it for a couple devices, most get a static IP.

It's worth noting that the other 5 scopes (set to 0.0.0.0) still work via DHCP relay from the switch. The server network scope (set to the server's own IP) is the only one failing.

The lease time for my server network is 30 days, so I hadn't noticed any issues. I have three DNS servers and wanted Technitium DNS to provide the DNS on my tailnet as well. This part is working.

However, since installing Tailscale, I'm no longer getting IPs assigned on this network, the other networks are fine. I stopped Tailscale, with tailscale down and stopped tailscaled, but it does not appear to re-enable handing out of IP addresses on the 90 network.

Setting static IPs on the couple machines did resolve the network issue. Although I have move a NAS VLANs give it a static IP and then move it back.

Any ideas besides removing tailscale completely from this server?

Could I force the listen address to 0.0.0.0?

[ShreyasZare]

Thanks for the post. If the scope range matches with the network interface on the server, then it will directly listen on the IP address on the network interface. In this mode, it wont support DHCP Relay Agent since its literally on the same Ethernet broadcast network.

I have no experience with Tailscale. Is the scope listening on the network interface that Tailscale provides? Does tailscale support broadcasting Ethernet packets?

[jelmervdm]

Sorry, I initially posted this hoping it would be a quick fix and I spent my weekend out in the desert instead of in front of my pc :). I understand that it's on the same network and relay shouldn't be necessary.

It is not handing out IPs on tailscale (although it does do DNS to the tailscale network). Tailscale is an overlay vpn, each endpoint authenticates to the management plane and gets assigned a /32 IP from the management plane in the vpn range when it first authenticates. By default 100.64.0.x. I have Technitium running in an LXC on proxmox and it does add routing rules to the instance, routing table 52 if I'm not mistaken. I was just wondering if you heard of it before or why it would be interfering with DHCP when it's on the same subnet, but not when it's listening on 0.0.0.0.

I'll do some more digging and see if I can find why it would not hand out IPs.

[ShreyasZare]

DHCP protocol works using broadcast UDP packets so the underlying network needs to support broadcast to be able to use DHCP. I could be that broadcast packets are not being transmitted over tailscale, not sure about it entirely though but that could be the reason why the IP allotment is not occurring. The other reason could be simply a firewall issue that is blocking the requests somewhere on the network.

You can check the DHCP logs on the Logs section on the panel to check if there are any DHCP requests coming from the clients. You can also try running tcpdump on the server and try to join a new client to the network to see if the DHCP requests are reaching the server. Both these things should make the issue much clear to understand. Do try and let me know what you see.