pxe boot – isc-dhcp-server (dhcpd) gives old next-server in DHCP offer

I’m running isc-dhcp-server (dhpcd) version 4.4.1 on ubuntu 20.04 with a subnet configured like so, which has been working just fine.

# /etc/dhcp/dhcpd.conf
subnet 10.0.1.0 netmask 255.255.255.0 {
  option routers 10.0.1.1;
  next-server 10.0.1.2;
  filename "pxelinux.0";
  allow bootp;
  allow booting;

  (..)
}

I recently tried changing next-server to another IP 10.0.2.2 and restarted the DHCP server, however when a PXE client issues a DHCP discover it responds with a DHCP offer that contains the old next-server IP 10.0.1.2.

Could it be that the bootstrap protocol (bootp) information linked to a lease? No reference to next-server in /var/lib/dhcp/dhcpd.leases from what I can see at least.

For what it’s worth I tried changing in another subnet and that works, but haven’t been able to identify any differences in their configuration.

networking – isc-dhcp-server host can’t reach the clients

I’ve setup an isc-dhcp-server on an odroid-h2+ on one of the LAN ports, while the other one connected to my main router. I have the dhcp server network port connected to a PoE router with one camera plugged in and my laptop is also plugging in on that side.

I can see that the isc-dhcp-server is leasing IPs fine (i’m using webmin to make management a little easier). From the laptop (on the same subnet), I can reach the web interface of the camera and can view the video stream with issues. When I try to navigate to the camera web interface from the odroid-h2+, which is running the dhcp server, I can’t hit any of the leased IPs, including the camera. Is there additional setup I need to do?

debian – Automatic restart of isc-dhcp-server

I’m a Debian and Raspbian newbie, and have a question about automatically restarting the ISC DHCP server.

I have a Pi running Raspbian and the ISC DHCP server at a remote site with flaky power. Everything’s on a UPS, but when the Pi comes back after power loss and UPS battery drain, systemctl shows the DHCP server status as failed.

I think this is an order-of-operations issue, where the Pi comes online before the router or switch are up, and can’t start the service as a result.

How to get systemd to handle automatic restart of the DHCP server when it gets in this state?

I’m aware of the “Restart=always” command for systemctl, but can’t find a unit file for the DHCP server to append it. The dhcp client has one, but not the server. This puzzles me, since the output of “systemctl list-unit –type=service” includes the DHCP server.

Thanks in advances for clues on getting this service to restart.

18.04 – isc-dhcp-server and systemd-networkd

I think systemd-NetworkD Support for DHCPServer is limited based on https://www.freedesktop.org/software/systemd/man/systemd.network.html#%5BDHCPServer%5D%20Section%20Options

However, you are probably better off with it ISC DHCP server If you need to tune multiple DHCP subnets simultaneously or provide advanced DHCP options and the like.

I have an LXD network that uses an 18.04 container that provides DHCP to all other containers on this subnet. It seems to work systemd-NetworkD and netplan DHCP clients.

Note that in either In any case, the DHCP server should have a static IP address outside the DHCP pool. This way, the server behaves consistently.

Can the isc-dhcp-server update bind run on a non-standard port?

I do the following on the same server: pi-hole, bind9, isc-dhcp-server, and cloudflared (for DNS over HTTPS). The order goes like this:

Hole: 53 -> Bind9: 54 -> Cloud: 5353

I'm now installing isc-dhcp-server as my authoritative DHCP server because my Ubiquity router can not really manage hostnames well and does not update DNS (uses a host file). I used to set up DDNS in the past, but now that I've run Pi-Hole on the standard DNS port (53), I'm not sure if dynamic updates are working properly.

I tried to look at the manual for ISC-dhcp-server, but all I can find for the primary DNS is an IP, not an option for the port. My question is, how does the isc-dhcp-server update bind? I know it uses a TSIG key for authentication, but does it still work if Bind is running on a different port?

Many Thanks!