Intermittent DHCP Problem, drving me mad!

[Zourous]

When the problem happens the PC fails to get an IP address and assigns itself a 169 address. Attempting to release and renew doesn't seem to work, but doing one of the following allows it to pick up an address again:

 

disconnect / reconnect the network cable

disable / enable network connection

restarting the PC (most of the time)

changing the speed of the card or making driver alterations

 

I've tried

 

Updating the network driver

Setting the network card speed permanently to 100mb

putting different images on the PC's

 

The network cards are 1 gb Broadcom net extreme 57xx

The problem seems worst in 2 or 3 rooms although I am seeing the same dhcp events on other PC's as well so I can't rule them out.

 

I have spotted in DNS that PC's in one of the troublesome rooms share IP's with other PC's. Could this cause this type of problem? If so, I'm not sure how it then works if you disconnect reconnect the network cable.

 

This problem is drving me mad because of it's intermittent nature so if anyone has any clues that would be much appreicated.

[apeo]

Seems an odd problem but (im sure you've checked this already but had to ask) have you checked to make sure the network is functioning properly. From the sounds of it I would guess its a network problem as its assigning fine in other rooms. Edited October 30, 2008 by apeo

[Domino]

have you checked that

1: you haven't run out of ip addresses in your range

2: there isn't a rogue DHCP server on the network

[matt40k]

We had a simpliar thing, can't remember exactly what it was.

 

I think it could be loopback, a rogue DHCP server, high network traffic, switch problem (fireware here we go! + physical setup)

[dtakias]

Can't think of something else rather than what is been posted already by the others

[Michael]

If you allocate the PC a temporary static IP does it work ok? The purpose of this is to identify whether it's a DHCP or a connectivity problem. If connectivity is still lost with a static IP, it would most likely be a switch problem.

[Face-Man]

s well as checking you have some spare addresses for you DHCP sever to issue I would check the age that they are issued for. You may have enough addresses in the range but they are taken up by old computers. I would clean up your DNS as well if you have multiple DNS records. Do you have some devices with static address inside your DHCP Scope?

 

If there are enough DHCP addresses your problem is probably a conectivity one. I would check for errors on your switches. Is there a router between you PC and the server ?

 

Have you tried assigning the PC a static address? Does it then work ok?

 

Good luck

[Zourous]

Thanks for the input so far. Just to update people:

 

The network is working normally as far as I can tell.

 

My boss recently added some more ip addresses so we should be ok.

 

I've checked for rogue DHCP's but can't find any.

 

It shouldn't be high network traffic as the kids are off and it's still happening

 

I imagine a static address will work fine, it's getting an DHCP address that is the problem. If it gets an address it doesn't then lose connectivity.

 

A PC can be fine working with an DHCP address. I shut down the PC and reboot and all of a sudden it can't contact the DHCP server. I have been testing it on around 22 PC's in a classroom. I shut them all down, fire them up and I will likely get 1 or 2 random PC's that refuse to get a address. To add to it, one's that have problems connecting sometimes come up with the "this windows is not geniune" routine. This disappears if they connect to an address the next time. I'm not sure if this is related or not.

[Domino]

could also be a DNS problem?

 

Anything in the event log after a machine has issues finding a DHCP server?

[Oops_my_bad]

depending on your switches, the switchports your PC's are connected to may be configured as "trunk" ports, ie they think they are uplinks to other switches and not connected to PC's so don't come up quick enough in time for the DHCP process to complete in time. Try changing them to "portfast" or similar if you can.

[Zourous]

Thanks for the help so far.

 

I've changed a setting on the switches which is 'state' to disabled under the spanning tree parameters. For some reason this seems to lets all PC's get a IP address. Some people had already mentioned the STP fast start but this was already enabled when I checked.

 

I'm not an expert on switches at all, so does anyone know what this really means? I understand that each port goes through different states (blocking, forwarding) when a PC turns on. Does this just turn this off? I'm scared of the implications it might introduce. I've read that the spanning tree protocol is there to protect again any network loops.

 

The PC's that have the problems are brand new and boot very fast, so I guess this is why other PC's on the same switch are not affected. The switch must be causing the DHCP to time out.

[Roger]

Have you looked at the DHCP logs? they can give you lots of info on what is going on, for example code 50 errors would be a rogue server on the network.

They can be found in C:\windows\system32\dhcp - you will see them listed by day.

We had a similar intermittant problem, ours turned out to be a replication issue between servers affecting DNS also.

Roger

[farmerste]

have you unticked the 'allow computer to switch off this device to save power'

box, that sorted out a similar problem for me,

or i would stick a 3com 3c905 nic in there for a while and see what happens, if the nic is integrated in to motherboard, try to flash the bios

 

hope that might help

[diggory]

If you have added a range of ip addresses to your DHCP server, they might not be used - if they are a new subnet. So you might still be running out of addresses (just a thought).

 

Worth right clicking on each scope/range on the DHCP server and getting the stats, to see if ip addresses are being used (or used up) - assuming you have more than one DHCP scope?

 

Oh yes I ought to mention, that 100Mb half duplex is worth trying - sometime you can get duplex errors between different NICs (only on 100Mb afaik)

[rvdmast]

Just out of curiousity, what brand/type switch do you use? Doesn't happen to be Dell, does it?

 

edit:

And the PC's ? Dell optiplex 755 by any chance?

Edited November 19, 2008 by rvdmast

[Zourous]

Thanks for all the input from everyone.

 

I think I've solved the problem now. I turned off the spanning tree protocol on all the ports and have had no problems since. This isn't the same option as SPT fast start. Obviously SPT is there to check for loops, but we haven't got any reduntant links so we should be ok so long as a student doesn't bend a network cable out of one datapoint to another.