DHCP Server in VM not consistent

[joebrug]

Last week, I disabled the DHCP server on my physical Netware box and installed/configured DHCP on two of my Windows 2008 virtual machine servers.  I have three Esx 4 hosts with latest version on them.

DHCP has never been an issue in the past, but since the change, randomly users will turn on their pc's and as windows starts to boot, the login screen never comes up, they just get a blue windows background forever.  Powering off the machines and back on (sometimes a number of times) seems to eventually fix the problem.  I've tried to ping the workstations while they're sitting at the background but am unable to.  So, I'm assuming they're not getting ip addresses properly from the DHCP server sometimes.

My question is, is there something in the network configuration that could potentially throw this off?

Each of my ESX servers have a quad port NIC card in them, all assigned to the virtual machine network.  With the two built-in NIC's hosting the Kernel and Service Console.

They're using defaults, as far as teaming, etc goes from what I can tell.

There are no errors on the DHCP servers (windows) themselves.  I'm wondering if the DHCP requests are either being unanswered, or they're going out on some different port than they came in on.. something weird like that.

Any ideas? Appreciate it

[ssumichrast]

We are just dealing with this in our environment migrating from Novell.  We reserve everyone's laptop in DHCP so they have the same address at our various sites (not many machines).  There's a fix out for 2008 R2 DHCP servers regarding reservations (it may also apply to dynamic leases) that you need to request the hotfix for (I don't know why they don't just release it as a regular update).

Anyways, check out this KB from Microsoft:  http://support.microsoft.com/kb/981776

It specifically is for multiple dhcp servers providing the same scopes (either split scope, synced reservations, etc).

I don't know for sure if it's related but it may be a start.

I've run numerous 2003 DHCP servers in VMs just fine (with or without reservations).

[mldmld]

Hi

I would install wireshark + winpcap on my dhcp server or on a VM connected to the same network to see the network traffic.

Wireshark capture network traffic and add comments, so it's easy to read the exchanges between servers

[joebrug]

Nice find.  Not sure it totally applies either but worth a shot.  I actually just turned off the dhcp service on the second server.. I guess that should tell me if that scenario applies, right?

I plan on doing a wireshark on the users desktops once they experience the issue.  Since the DHCP server logs don't show any errors or anything odd at all, I'm not sure the request is getting that far.

Was more wondering if theres any configuration as far as the nic's go on the vmware host side that I might be overlooking.  Like maybe the request comes in to one NIC port, and the VM's respond and send it out another port, confusing the client and it just hangs.. grasping at straws here really..

Thanks for the ideas

[ssumichrast]

I don't think you're seeing a VMware issue.  It shouldn't matter which vNic you come out of as long as both vNics have access to the same layer 2 domain it came in on.

I'd probably next drop Wireshark on the DC as someone else suggested above.  One symptom we saw regarding the hotfix I posted above is that DHCP requests flood the DHCP servers because one is returning invalid data.

Try disabling the scope on one of the DHCP servers and see if it fires up.  If not reverse it.  If it fires up, I bet you're somehow affected by the above hotfix.

If not, make sure your IP helpers are updated (if applicable).  Wireshark should really provide some great insight here.

[joebrug]

I was actually on the phone with Vmware support earlier just to verify that my network configuration was correct, they stated there didnt seem to be any issues.. so thats good.

Correct me if I'm wrong, but stopping the DHCP service on Server#2 should be good enough to see if theres a conflict between server1 and server2 right? Unless server1 is the problem, of course.  Just making sure I wouldnt have to still disable scopes, etc if I'm turning the service off

[RParker]

Something else is wrong with your configuratoin on workstations.  I don't care what network configuration you are using a workstation should NEVER hang because of LACK of IP.  you have something else wrong..

It either gets an IP or it doesn't, Windows has NO requirement for IP related services.. so you can blame DHCP, but that is NOT the issue.

[joebrug]

I agree.  The login page should still come up regardless.  My assumption that it was DHCP related is based on the fact that I cant ping the workstations, etc in that 'hung' state, and since thats the only thing thats changed lately.  Was wondering if it was hanging during the dhcp process instead of the "yes you can have an IP" or "No you cant" ... maybe it is getting stuck in limbo.  Not sure what else to look at in the meantime..