6

I have a Server 2008 R2 DHCP Failover Cluster that serves two VLANs - an office network and a guest WiFi network. Both of the members of the failover cluster are virtual machines (ESXi) and use the iSCSI Initator inside the VM for their shared storage. Each node in the cluster also runs other FO Cluster services that are working fine.

However, with the DHCP service, in the past 3 weeks it has started only ever remembering one of its DHCP address offered. This is a real problem because it basically starts offering in-use addresses causing IP conflicts all over the network. Every now and again it will go and fill up its table with "BAD_ADDRESS" entries to cover the in-use IP's that it finds, but then they go away as well.

This is what the scope leases look like at the moment. There's only one non-reserved lease listed: enter image description here

When I connect my iPhone to the network, it takes literally about 45 seconds to get an IP address:

enter image description here

And when it does, it's not showing up in the lease list anywhere. Which means that next time someone else connects a new device, they will end up with the same IP address and then a BAD_ADDRESS will be marked in the list - until the next device connects when the whole thing will be lost and it will go and re-mark them all:

enter image description here

The only thing I can think of is that these cluster members used to be Hyper-V members and were migrated to ESXi at about the same time these issues started occuring - but I can't be 100% sure.

7
  • Have you looked at the network traffic on your DNS servers to see if they're actually issuing any IPs? Weird delays, conflicts and missing leases suggest a rogue DNS server to me.
    – DerfK
    Apr 17, 2013 at 23:44
  • @DerfK - Honestly, I haven't gone in that deep yet. My next step was actually going to be to kill the cluster and set it up standalone, or maybe two servers in an 80/20 split. Apr 17, 2013 at 23:52
  • For what it's worth Mark, I've used split scope DHCP for every DHCP deployment I've done and haven't had any issues. W2K8R2 even has a split-scope wizard you can use to configure the scope.
    – joeqwerty
    Apr 18, 2013 at 0:00
  • Is the DHCP server running as a ESX VM? If so check it's date/time settings compared the clients it's serving. I don't know how well ESX and Windows get along but it's terrible with linux for time drift. If the time gets too off then the lease is gone based upon the lease time frame. If there's time drift try increasing the lease time.
    – John
    Apr 18, 2013 at 0:00
  • 1
    @ashley - I was sick of maintaining two seperate environments so everything got rolled into ESXi and vSphere, with VMWare site recovery manager to be added into the mix in the next phase. Apr 26, 2013 at 11:08

1 Answer 1

6

Enabling Conflict detection (which is not enabled by default) will probably resolve this.

3
  • This actually seems to be making things a lot better - I now have 10 valid leases and three BAD_ADDRESS's Apr 18, 2013 at 0:24
  • 1
    Although I'm unsure why enabling conflict detection meant that the DHCP server now remembers all its previous leases, it would appear that it does. (and we no longer get dupes from previously allocated addresses). I now have a large list of DHCP clients, and its growing every few hours. Thanks again. Apr 18, 2013 at 3:22
  • Glad to help...
    – joeqwerty
    Apr 18, 2013 at 3:24

You must log in to answer this question.

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