7

My network was working just fine and I broke it.

There is a Cisco 1711 router which was configured to hand out DHCP leases to the client PCs. I have a domain controller (Server 2008) on the same network that I wanted to take over the DHCP role. I added the DHCP role to the server, configured the scope, and activated it. Then I turned off DHCP on the Cisco 1711 by running no service dhcp in config mode.

I tried to renew the IP address on a client, and it failed. I ran Wireshark on the server, and it sees the three DHCP Requests that the client sends before reporting a failure. Absolutely no DHCP packets are sent in response. In the DHCP mmc, Statistics tells the same story:

Discovers        0
Offers           0
Requests         6
Acks             0
Nacks            0
Declines         0
Releases         0
Total Scopes     1
Total Addresses  50

The DHCP service is running and authorized. I checked C:\Windows\System32\dhcp\DhcpSrvLog-Mon.log to confirm this.

00,01/24/11,13:46:23,Started,,,,,0,6,,,
55,01/24/11,13:46:25,Authorized(servicing),,cfl.local,,,0,6,,,

I've triple-checked my scope configuration, it's definitely correct. The server is not multi-homed. I disabled and re-enabled the DHCPv4 bindings on the single network adapter. I even restarted the server. I can't think of any reason the server would just ignore DHCP requests.

It gets weirder. I decided to give up on using the server and return to using the Cisco router as the DHCP server. I ran service dhcp in configuration mode to resume the DHCP service. I tried renewing the DHCP lease on the client PC again, and it STILL failed. Then I turned on debugging (debug ip dhcp server events and debug ip dhcp server packet) and tried renewing DHCP on the client again. There was absolutely no output to my SSH session with the router! It's as if the packets are disappearing completely.

The network is simple and flat. All of the devices (PCs, server, router) are connected to the same Layer 2 gigabit switch.

Nic
  • 13,025
  • 16
  • 59
  • 102

1 Answers1

4

You haven't actually verified that the Windows Server machine is receiving the DHCPDISCOVER requests. I'd throw a sniffer on it (either Microsoft Network Monitor or Wireshark) and verify that the packets really are making it to the server computer.

Are you certain your Ethernet switch doesn't have any "fancy" DHCP-related functionality that might be getting in the way? Your disappearing DHCP packets that you describe toward the end of your posting make me wonder if the switch is "swallowing" them somehow.

Edit: Zow-- I'm a moron! Talk about not being able to read.

Here's a silly question: Is the server computer's NIC configured with an IP address inside the subnet that the DHCP scope encompasses? If not, you'll need a DHCP relay agent (which could be running on your router) to forward the request to the server because, even though it's receiving the request at layer 2, if the server doesn't have a NIC in the subnet with the scope (or if the request doesn't come from a relay agent in the subnet that corresponds to the scope) the server won't answer.

If they're just RFC1918 IP addresses, go ahead and edit your question to include the server computer's IP address and the DHCP scope metes and bounds.

Evan Anderson
  • 141,071
  • 19
  • 191
  • 328
  • I did run Wireshark on the server and the server is definitely receiving 3 "DHCP Request" packets but not "DHCP Discover". – Nic Jan 24 '11 at 22:17
  • @Evan: I think you mean "in the same subnet" not "in the same scope" in the first sentence of your edit. – joeqwerty Jan 24 '11 at 22:23
  • 1
    @joeqwerty: This shows why I shouldn't sneak over to Server Fault while waiting for a reboot to answer a couple of questions quickly... >smile – Evan Anderson Jan 24 '11 at 22:24
  • Speaking of "I'm a moron" ... I had configured the address pool range to be 192.168.125.50 - 192.168.125.100, but the server address was 192.168.125.1. I expanded the address pool and it works now. – Nic Jan 24 '11 at 22:25
  • 1
    What subnet mask is in use because the original range looks fine to me for a /24 subnet. – joeqwerty Jan 24 '11 at 22:26
  • Joe, the subnet mask is 255.255.255.0, and that's what I thought too. But apparently that isn't what Windows Server 2008 thinks. – Nic Jan 24 '11 at 22:28
  • That's weird. So what is the new range? – joeqwerty Jan 24 '11 at 22:29
  • The new range is 192.168.125.1 - 192.168.125.254. I haven't configured any exclusions yet, I'm just happy that it's working. – Nic Jan 24 '11 at 22:30
  • Definitely a strange one. – joeqwerty Jan 24 '11 at 22:31
  • Definitely odd. The W2K8 DHCP should've had no problem serving addresses in a scope with the bounds 192.168.125.50 - 192.168.125-100 with the server's NIC being set at 192.168.125.1/24, assuming the scope's subnet mask was also set to /24. Oddness. Glad it's working for you now. – Evan Anderson Jan 24 '11 at 22:35
  • Thanks for this answer. I had configured the subnet mask incorrectly on the scope, causing the server's IP to be outside the subnet. Once I fixed the netmask it started working correctly. – Greg Bray Jan 02 '13 at 06:07