DHCP secondary server is configured with 20% of address leases and if it serves the clients at the same pace it would run out of address leases and would be useless if the primary DHCP server is down. To avoid address range depletion on the secondary server one of the following two methods can be employed
· A delay is set on the RA (Relay Agent) in the proximity of the Secondary DHCP server. A delayed OFFER from the secondary DHCP server will make the client to request the lease from the primary DHCP server and will only request from secondary only the case where primary is not reachable. Setting the delay in the RA ensures that the client receives the OFFER later than it receives from the primary, but it does not delay the OFFER originating from the secondary server itself. In this kind of setup the RA should be setup mandatorily even if it is not required.
· Delay is set on the secondary Windows Server 2008 R2 DHCP server using MMC/Netsh tools. This would make the OFFER originating from the DHCP server is delayed. In this setup the RA is not mandatory.
Configuration Steps (set delay on secondary DHCP server):
Launch the scope properties and change the Delay configuration value in the Advanced tab of the scope in particular:
This can be configured also while configuring the split-scope using the split-scope configuration wizard. (refer the blog on wizard based split-scope configuration)
[in, unique, string] DHCP_SRV_HANDLE ServerIpAddress,
[in] DHCP_IP_ADDRESS SubnetAddress,
[in] USHORT TimeDelayInMilliseconds = 0
[in] USHORT TimeDelayInMilliseconds
Hope this helps in managing your addresses. Comments, suggestions or queries are welcome.
I am interested in how this might work instead of 80/20 rule type setup and basically in a 100/100 setup between the Primary and Secondary server. In theory, it seems would work the same. However, I am interested in possible caveats that could result from this? Would it even work?
Also, does Microsoft have plans to change up the way DHCP is delievered? Perhaps moving from the current "single instance" Jet database to something that is "known by all participating DHCP" servers would be much nicer. Maybe utilizing ADLS or even moving this to similar design as DNS the replicates in its own AD partition? We are facing more pressures to offer a HA/Redundant solution that will require 'little interaction' in teh event of a failure. Products which our network team looks at are Cisco's CNR and things like InfoBlox appliances.
Sr. Systems Support Specialist
Global IT Infrastructure
I'm having this exact problem, however our DHCP is set up on two 2008 Enterprise servers, not R2. We have several split scopes working, except one. In this particular scope, the secondary (70/30) is leasing IP's and we're running out. When we reach our limit, server 1 isn't picking up the slack. First, why isn't server 1 providing addresses first, and how can we avoid running out of addresses? I know we could add more, but that just seems to mask the problem. Can we use configuration features described in R2 manually, somehow? We do not have any exclusions set up.
Davidson County Community College
What's the value for the delay configuration?
Lets say you have 500 IP's being handed out by 2 DHCP servers runing split 80/20 scopes. During the morning as clients are booted up and doing DHCP discovery, the 80% client goes offline for whatever reason. This would force all clients to get IP's from the 20% scope and would exhaust all IPs frmo the 20% scope. How do you now force those clients to renew their IPs from the 80% and not the 20% without stopping the DHCP Server Service on the 20% server?
Also, how do you handle reservations? With 2 servers, you now have to add the reservation on both DHCP servers?
I'm starting to think that split scope is added complexity that is not needed.
mallettero: did you managed you problem with 2008 servers? we have the same problem here