How can I configure a DHCP server to assign addresses based on the OS that is running Solved maybe!

[Chris Buxton]

On Jan 17, 2011, at 1:54 AM, Niall O'Reilly wrote:
> On 16 Jan 2011, at 17:29, Marc Chamberlin wrote:
> 
>> But I do have a bit of reservation, I think this approach would be OK in a SOHO or small business environment, but not so sure if it would be a good idea in a larger enterprise environment with hundreds of users. 
> 
> 	With dozens of networks, perhaps twice as many lease pools, and thousands of students
> 	using laptops and smartphones, we don't see problems with 10-minute lease times.
> 	We do see "no free leases" messages, but that's a different problem.

I've seen issues with several tens of thousands of clients hitting a single DHCP server, using 30 minute lease times. The issue was primarily to do with the speed of ddns updates -- a little hacking of the BIND source code fixed that (commented out one line to make file writes asynchronous).

For that environment, we saw something like the same problem as the OP, but we weren't doing anything as complex as the OP posits. We were able to get away with just using one-lease-per-client and modifying the hostname for wireless clients (adding -wifi to each hostname on wireless segments), so that a client's wired and wireless NICs can both have leases at the same time and both have unique hostnames. I don't see that working here, because the OP wants the same IP address between two clients (the wired and wireless NICs), but different between two instances of the same client (wired NIC, two different operating systems). It seems convoluted.

Perhaps a better approach would be to use ddns, and have the backup system use DNS lookups. Also configure it to try wired first, then wireless, using a formula for calculating the wireless hostname from the standard hostname (e.g. the -wifi trick described above).

Regards,
Chris Buxton
BlueCat Networks