Recently I've created my own thin client operating system (Xubuntu based, so basically Linux yes) that will run across our thin clients. Currently I'm testing the OS in my test environment, but I've come across the following problem.

Our production environments will assign them DHCP private IP-Addresses, and add the client in the domain using SAMBA4. They are now reachable based on IP, but since it's DHCP, the IP will change basically every two weeks. It will be too hard for me to note down all IP's every week and document them. So the solution will probably be DNS, but I'm quite unsure what is the most efficient way?

I've thought about the following solutions:

  • Place the hosts in a special subnet, give them static ip addresses and create A record in DNS.
  • Why do I want to reach it hostname?
  • I want to UltraVNC using a standard hostname, and not every 2 weeks having to change another IP.
  • I will be pushing changes with Ansible, I prefer to have a standard hostname, instead of a random IP.
  • Are we talking public IP addresses or private IP addresses? (Please edit question). DHCP can also hand out static IPs based on machine information (machine name in DHCP request, MAC address), that would be the simplest solution. Then you can use static DNS records.
    – dirkt
    Jul 25, 2018 at 9:09
  • Private ip addresses. im sorry will edit it.
    – David Deng
    Jul 25, 2018 at 9:56

4 Answers 4


I would use a mix of strategies.

Define static IP addresses for the network/netblock of desktop PCs where people give you the MAC address and you register the name.

Define dynamic IP addresses for BYOD networks, such as wifi networks.

The added advantage of having static IP addresses is to be able to pick up logs a few months old, and not have to cross-reference them with DHCP logs to find out what machine you are dealing with.

Another approach might be @roaima solution of DDNS, however in that case I would use far more than 1-2 weeks time for the lease time.


Dynamic DNS is what you need. The client updates its own A record each time it changes. (This isn't a paid Dynamic DNS service you can get for DHCP-based hosts on the Internet, but it's the same principle.) The bind DNS server can handle DDNS for a zone; dnsmasq also will do it.

Domain-joined Windows clients have used DDNS automatically with AD for a long time. I now see this with recent domain-joined Linux clients against Windows AD, but I haven't tested against a SAMBA AD.

You should find that SAMBA will provide this service automatically - see https://wiki.samba.org/index.php/Samba_Internal_DNS_Back_End#Setting_up_Dynamic_DNS_Updates_Using_Kerberos for what appears to be a statement that it is supported through the allow dns updates parameter in the server's smb.conf file. However, as I've already mentioned, I don't run SAMBA as an AD domain controller so I have not tested this.

  • Cool, haven't heard about that one. I've haven't heard about it before. I guess its a paid service? What do you think about this DDNS provider? noip.com
    – David Deng
    Jul 25, 2018 at 8:01
  • is hosting a own dynamic dns server also an idea?
    – David Deng
    Jul 25, 2018 at 8:14
  • @DavidDeng answer updated. I hope that's clearer for you. Jul 25, 2018 at 18:50

If it's private IP address, I'd definitely assign static leases. In ISC dhpcd, for example, you can do that with the fixed-address keyword, matching e.g. on the hardware ethernet (i.e. MAC) address.

ISC dhcp can also dynamically update your DNS server, if you still need that after assigning static leases. The keyword hostname will set the host name.

Having each client dynamically update the DNS server is overkill IMHO; this makes sense for public IPs when the ISP assigns the address and you have no control over it, but not for private IPs.

  • hmm sounds reasonable, currently our endpoints(mobile devices , laptops, thin clients) are all DHCP in a scope of 254 hosts. Do you think its a better practice to move the new thin clients all into a different subnet? so i only get 1 subnet of static leases hosts ? or should i keep them in the current dhcp scope
    – David Deng
    Jul 25, 2018 at 13:50
  • Different ranges for static and dynamic leases are definitely a good practice. They don't have to be on different subnets, that's an orthogonal issue. Subnets are determined by your network infrastructure. And you can have subnets larger than /24 (i.e. more than 254 hosts).
    – dirkt
    Jul 25, 2018 at 14:05

I have a similar setup, where I have an AD domain (running on Windows 2008) but want to manage the network with a Linux box. I have set up dnsmasq so that it's the authoritative DHCP server for the network and the authoritative DNS server for the AD zone, forwarding the infrastructure DNS queries to the AD server; dnsmasq then automatically maintains the DNS entries.

Here's a redacted version of my dnsmasq.conf (with most comments removed):


# Refer infrastructure DNS requests to an AD server






The /etc/dnsmasq/hosts contains any hosts that are needed to contact to AD servers (or use static IP addresses): ad1
2001:db8:1234::254 ad1 ad2
2001:db8:1234::253 ad2 router
2001:db8:1234::1 router dns
2001:db8:1234::2 dns

You don't need to change any DNS settings on the AD server, but you may want to set the forwarder to the dnsmasq box (the AD server will work in split-horizon mode locally, which is fine.)

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .