2

I'm running two DHCP servers both with Windows Server 2012. They are set up as with high availability using "load balanced, split 50/50." They named DC1 and DC2.

This morning, DC1 was frozen, there was a black screen and the computer was completely unresponsive. There is nothing in the event log that suggests what went wrong. I ended up having to hard reset it. It seemed that DC2 was not picking up the slack while DC1 was down. I noticed because a new computer I plugged into the network was not being assigned an IP address. Also, address renewal had failed on another computer, this was in the event log:

Your computer was not able to renew its address from the network (from the DHCP Server) for the Network Card with network address XXX

Your computer has lost the lease to its IP address 192.168.1.XXX on the Network Card with network address XXX.

If DC1 is not working, shouldn't DC2 automatically start issuing IPs and addressing renewal requests? Or is this something that has to be configured somewhere? I'm not very familiar with how this works, so any assistance would be appreciated.

I'm wondering if this was because DC1 wasn't actually unplugged or turned off, just frozen?

5
  • Can you confirm that Failover is configured on the scope in question? DHCP Failover only requires a server to fail responses of a DHCP request before the secondary would pick it up so a frozen/locked system should still trigger it.
    – Byron C.
    Jul 21, 2014 at 23:57
  • Indeed, if I right-click the scope in the DHCP window and select Properties > Failover, the correct relationship is listed. The current mode is "Load balance." Jul 22, 2014 at 0:01
  • If DC1 ever locks up again, I'll know to check on DC2 to see what the failover relationship looks like. Right now both have their status listed as Normal. Would have been useful to see what that was earlier today when DC1 was frozen. Jul 22, 2014 at 0:02
  • Check the scope statistics. That can show if ANY addresses are being served by the secondary pool. If not, your best option would be to do some packet capturing. Depending on your network topology, you may need to work with the switches/routers.In our environment, this issue turned out to be related to DHCP helper address misconfiguration.
    – Byron C.
    Jul 22, 2014 at 15:45
  • Just as a reminder, DHCP will by default wait a long time (like 1 hour or so) until failover happens. You can decrease the value.
    – Daniel
    Jul 22, 2014 at 20:55

2 Answers 2

1

If DC1 is not working, shouldn't DC2 automatically start issuing IPs and addressing renewal requests?

Yes. In a Load Balanced DHCP Failover relationship the DHCP partner server will assign a DHCP lease for its unavailable partner.

How it works:

DHCP Failover partnerships create a hash of the client MAC address when a lease request is recieved. Both servers will know which hashes it needs to respond to and both servers should be able to see requests. If a request goes unanswered after a handful of tries from the client, the partner server will respond with a short term lease from its IP pool equal to the MCLT length. (So it can expire the lease quickly when the unavailable server is back up). Server 2012 Reference and Obscure DHCP Failover ietf Reference

DHCP requests come about 1 second apart and this entire process usually isn't noticeably longer (to a human) than receiving a standard DHCP lease.

is this something that has to be configured somewhere?

You should of course check your configuration if you are not familiar with it. Going through a good how-to would be a good way to check your configuration. You can also do a quick check on a scope from the DHCP Snap-In by Right-Clicking the scope and selecting Display Statistics. This will show how many IP addresses are assigned and available from each failover partner.

DHCP Failover Scope Statistics

You can likewise view your partnership state by viewing the scopes properties and selecting the Failover tab.

DHCP Failover Scope Properties

I'm wondering if this was because DC1 wasn't actually unplugged or turned off, just frozen?

All these states will trigger the same Communtication-Interrupted state on the partner server. Your secondary partner should take over unanswered leasing regardless of why it cannot communicate with its partner.

Further Troubleshooting:

If your secondary server is still not able to serve leases in Communication-Interrupted state, your best bet is working with your network administrator. If you are he, I would suggest packet capturing DHCP traffic and checking for DHCP Helper Addresses corresponding to both DHCP servers on all intermediary switches.

-3

I've never had any luck getting DHCP load balancing to work on 2012 yet. Actually a lot of stuff on 2012 is really buggy and I've downgraded to 2008 R2 a bunch of times already. Until they get everything fixed. What you can do is disable the load balancing and just create a single scope on each DHCP server and make sure the two scopes don't over lap. If you don't have enough IP's in the split scope to serve the entire network, create the scope on a bigger subnet i.e. 255.255.254.0 for both scopes.

2
  • DHCP failover is working and be cautious with a statement like: "a lot of stuff on 2012 is really buggy". You need to back this up with evidence.
    – Daniel
    Jul 22, 2014 at 20:59
  • My personal experience is what I back up that statement with. Having done around a dozen 2012/R2 installs and every time something basic doesn't work right. Takes me back to early 2003/2008 days. Every time MS releases a new version things are buggy I'm not the only one that thinks that.
    – George
    Jul 22, 2014 at 23:32

You must log in to answer this question.

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