Possible DHCP Problem

[brahma]

Hello there - I'm new to this forum so am hoping you guys can help me out of a hole!!

 

I've inherited a Windows 2003 Active Directory domain that has got a single DHCP server on the network, about 20 HP Switches and about three hundred client pcs. The previous IT Manager set it up with a subnet mask of 255.255.252.0 and a single scope of 10.12.12.2 - 10.12.15.44 with a lease period of 14 days.

 

What happens is every 7 days or so, a number of PCs (that are all connected to the same switch) do their job and go and seek out the DHCP server to verify their address and to renew but they all lose their network connections - my helpdesk phone rings off its hook with users complaining! The other IT staff have been told that when this happens the "quick way" of fixing it is to reset the switch that all the PCs are connected to and this does fix the problem. When I check out the event logs on the affected PCs they all report issues with renewing IP addresses.

 

I have not seen this problem before and I have not seen a single scope setup in this way before. What do you guys think? I think I should have at least one other DHCP server on the network for redundancy anyway. Should I have multiple scopes setup for four network segments (e.g. 10.12.12.x 10.12.13.x 10.12.14.x and 10.12.15.x) on the one server?

 

Sorry for the long post but I'm stuck!!!!

[Whitestar]

Well. That scope is Ok.

That subnet mask allows for those host address ranges.

 

The fact that you have been told that you have to reset the switch to get the clients working tells me that you should be looking at problems with the switch.

 

If you can, try swapping that switch with another one and see what happens. If you find that the workstations start getting addresses and the other ones dont, you know the problem is the switch.

[FN-GM]

I agree it looks like a hardware problem. Does it happen to the same computers or random ones.

We had something similar once when the onboard network card mac address where the same as other computers on the network.

[joe90bass]

Yep, I'd start with the switch. What switch is it? It may be a known issue - Have you looked for a firmware update? Or it could be just plain hardware failure.......

[Michael]

I agree it looks like a hardware problem. I'd also check that nothing else is using DHCP's ports 67 and 68 on the network.

[brahma]

Thanks for the quick reply Whitestar.

 

The reason I thought it might be DHCP server related is I have a number of switches and this problem has happened on quite a few of them. When I check the event logs on the switches there is nothing recorded - they just report system up time (expect of course when the switch is reset!) In the past we have even had the switch management software open when the "issue" occurs and nothing is reported in the logs.

 

The client PCs all record DHCP problems - and they all occur at 50% through the lease period - it just seemed to much of a coincidence that the switch goes down when the client are trying to renew?

[Whitestar]

Try this.

 

Turn off all the affected machines but one. Then manually force a DHCP renew.

 

Sounds like the switch might be stopping what it thinks is a broadcast storm.

[brahma]

Ok Whitestar - I'll give it a whirl tomorrow.

 

Thanks for your help so far. Any ideas what I should do if it works / fails?

[mrtechsystems]

Also what the is make and model of the switch

 

Might be able to shed some light on it

[Michael]

Sounds like the switch might be stopping what it thinks is a broadcast storm.

 

You might be onto something there, but I've never known a switch to treat DHCP requests as a broadcast storm. It's probably more likely one of the machines has a faulty network adapter and is creating the broadcast storm.

 

I've seen it happen before on HP switches and all LEDs are on permanently.

[brahma]

Morning all.

 

The switches are all HPs - 2600's, 2800's and 5300XL's. Thanks for your help with this guys - I'm dreading the next DHCP refresh as I'm sure the same thing is gonna happen again!

[Geoff]

Does your DHCP scavenging interval match your extended lease time of 14 days? As the default is 7....

[brahma]

Hi Geoff.

 

Not sure what the DHCP scavenging is set to and I can not remember how to find out!!

 

DNS is setup to scavenge stale resource records every 7 days.

[Geoff]

Sorry I meant DNS scavenging. You must set your scavenging to be the same length of time as your leases otherwise scavenging will remove DNS records to machines when they are still active and in use.