DHCP Issue?

[Gongalong]

Hi folks,

 

We've done two bits of work this week. Replacing our edge switches (old 3Com 100Mb switches to new HP Gb switches) and (Windows) updating all our servers.

 

Since yesterday I've noticed that DHCP is behaving peculiarly. When running a Ghost session yesterday all the clients had APIPA addresses. IPCONFIG /RENEW got a correct address. Again today I've spotted some PCs with APIPA addresses, yet other PCs are fine (perhaps because they were left long enough to get an address).

 

There are a couple of errors on the DHCP server (Win 2012) re. contacting Active Directory, but there are one or two of those dating back to November as well.

 

Any ideas where to start trying to troubleshoot this? Could a Windows Update have triggered this? Could it be our new edge switches? (I've logged this issue with the consultant that installed the switches)

 

TIA

[ChrisH]

You should mirror the port of the client or use a lan tap and capture the traffic using Wireshark and look for time outs or rogue DHCP servers regarding the DHCP process.

[ricki]

Check to see if you have used all your dhcp addresses.

 

Restart the dhcp service and release and renew a client

 

You might have to reduce the lease time

 

Or you might have a switch thats misbehaving. Try restarting all your switchs

 

Have a look at the event log on the dhcp server and see if you are getting errors. You might need to update the driver of the nic card on the server.

 

You need to sniff the traffic of the network to see if a computer has a broken nic and is arping or broadcasting too much

 

Also make sure tree spanning is on on the switches just in case you have a loop

[Gongalong]

We have a scope of about 1,000 addresses, and this is half term, so it seems unlikely they're all used up. The problem seems to be that some machines are slow to get an address.

 

I've rebooted the DHCP server, just in case.

 

My guess is that it's the switches, as that's the other major change in the last 2-3 days.

 

DHCP server log seems OK, well about as OK as servers ever are.

 

The server is a VM, so uses a virtual NIC. I'll check to see if integration services have updated.

 

We did have a problem with STP, so it was reduced to a compatible version with our older core switch.

[FN-GM]

The problem seems to be that some machines are slow to get an address.

 

Ah ha! This is key. On your client switch port enable Port Fast. On a Cisco Switch CLI its: Spanning-tree Portfast

 

I am confident when you enable that on the ports for your clients DHCP issue will be resolved.

[Gongalong]

Ah yes, Google turns up lots on this! I'll speak to our switch guys... thanks!

[Gongalong]

(Presumably this would apply for HP switches also?)

[ChrisH]

Portfast is usually the default on HP.

[ChrisH]

Show spanning-tree should tell you what they are set to.

[Gongalong]

I know that Spanning Tree caused some problems when the edge switches were installed as we didn't upgrade our core (3Com) fibre switch (it's already Gb). There was a version conflict apparently which caused ports to shutdown, so an earlier version of Spanning Tree was forced on the HP switches.

 

We're relying on a thirdline consultant for this as switches are beyond my skillset.