In Part 1, I provided with walk through of how to install DHCP and configure scope on your newly installed DHCP server. In this post, we are going to setup failover between two servers. I am assuming that two DHCP servers are installed and a scope is configure already.
So let’s go ahead and configure failover for given scope:
In the resulting dialog, you can either select all scopes that exist or select scopes you want to provide failover for:
On next page, you need to provide partner server. This is the server where scope information will be replicated to:
After selecting partner server, you need to configure failover relationship parameters (you can read more detailed description on TechNet article: http://technet.microsoft.com/en-us/library/dn338985.aspx):
- Relationship name is descriptive and can be anything that helps you identify the relationship
- Maximum client lead time defines how much time extension a partner server can provide to a client, based on time known by partner server
- Mode allows you to choose between hot standby mode and load balance mode. In hot standby mode only one server services DHCP requests from clients. Load balance mode allows both servers in partnership to serve DHCP clients
- If you select load balance mode, you can define how each server will serve DHCP clients by defining percentage per server
- State switchover interval controls how long standby server waits after losing communication with primary server before assuming active state and servicing DHCP clients. In load balance mode it may not be important factor as both servers can serve clients based on defined percentage of scope IP addresses. In Hot standby mode, it will be important to select, or the standby server will never switchover automatically
- To secure communications between DHCP partner servers, you have an option to enable message authentication between servers. If enabled, you must specify shared secret.
After configuring failover mode and parameters, you are provided with summary page:
When you click finish, you will see progress dialog, informing you of status of failover configuration steps:
If all steps are successful, you now have a working DHCP failover pair, with no clustering required! Cheers!
I’m looking to implement DHCP Failover in my environment on two DC’s running Server 2012 R2 in load balance mode (50%/50%)
I’ve read numerous articles and still don’t fully understand the difference between the settings:
– “Max Client Lead Time” and
– “State Switch Over Interval”
I was hoping you could explain them in laymen’s terms and also recommend a “Best Practice” approach.
My thought was to set:
– Max Client Lead Time = 5 minutes
– State Switchover Internal = 30 minutes
My goal is to ensure that if one of the DHPC servers fails:
– the remaining server takes on the entire scope fairly quickly
– the load balance scenario comes back fairly quickly once the original server comes back online
– I don’t want failover to happen if I have to simply reboot one of the DHCP servers and it’s offline for a brief period of time (say 15-20 minutes)
Thanks in advance for any help you can provide.
Great question Rob!
Max Client Lead Time (MCLT) is defined in RFC for DHCP failover document: http://tools.ietf.org/html/draft-ietf-dhc-failover-12#page-7. There are guidelines on setting MCLT values in the document as well: http://tools.ietf.org/html/draft-ietf-dhc-failover-12#page-43. And as it usually is in anything IT, the answer is always “it depends” on the factors that affect how different values affect network and DHCP partner. You will have to decide for your environment whether short MCLT will work or should you prefer long MCLT values.
Sorry I can’t prescribe values for your goals, but hopefully the “light” reading I have mentioned will help you in either meeting your goal setting appropriate MCLT, falling asleep while reading RFC document, or possibly both! 🙂
hi, thank you for posting this very helpful information. I just have one concern. Is there a way to setup virtual IP for these 2 servers. In my environment, we are using ip helper address pointing at our primary dhcp server. In case the primary goes down, the standby server must have the same IP address as the primary. Is there a way to do these?
Hi, depending on your device where you configured ip helper, you should be able to add multiple ip helper addresses. I also came across an interesting post on cisco which may be relevant. https://supportforums.cisco.com/discussion/11327531/configure-cisco-router-ip-helper-redundantfailover-ms-dhcp-servers
Hello and thanks for this great post.
i would also like to raise some questions here!
My scenario is a failover configuration in Hot-Standby Mode.
Due to the environment of the clients and the complexity of the scopes (some have policies for chunks of ip ranges) coupled with the fact that we need to maintain a low lease time (3 hours) i have the following doubts:
1) Are there any implications by configuring a low MCLT (e.g 1 min) + 1 min for the Auto Transition Timer in order to achieve faster failover ?
2)what happens if we configure the standby dhcp server for 0% percentage of the scope? Does this mean that in case it becomes the active server (following a partner down event) it will not be capable of assigning new leases? Or does it mean that whoever is the active server has the full “control” of the scope?
Sorry I can’t really help with MCLT. Please read my response to Rob above. There are factors to consider and are unique to each deployment’s goals. Setting standby server scope to 0% is counterintuitive, don’t you think? What will it really achieve?!
Does configuring DHCP failover have any implications on DNS?
The obvious one, is when the IP address changes for the endpoint, the DNS registration will also change if the endpoint is configured to update DNS. IF an endpoint is receiving IP from DHCP, the shouldn’t be a reason to not allow dynamic DNS registration on that endpoint. Hope it helps.
I choose Hot Standby instead of Load Balance for MS Patching reasoning. In other words patching a Windows server sometimes takes a fair amount of time and a reboot.
1) The lease is set to 3 days.
2) The State Switch Over is set to 60 minutes
3) The Maximum Client Wait Time is 1 hour
So in this scenario the client should still maintain their lease while doing maintenance and will not switch over for 60 minutes. So the worse case scenario is if a new client requests DHCP or their lease expires is one hour. WHich in theory I could reduce but I think it is minor.
Your approach is certainly logical and should work without any issues I can foresee.