Spanning tree and DHCP

[John Anvari]

I have mixed networks of HP and Cisco switches. I also have networks of all HP Procurve switches. I use dhcp on all of the networks being serverd by the local server to the building.Having spanning tree on pretty much makes dhcp useless. So I had to turn it off on all of my switches. Of course a couple of "students" have helped me out by looping my network a couple of times. I have read the HP's rstp does not have the timeout issue for dhcp. Cisco of course does not have this option. What are the implications of some switches having rstp on (HP) and some switches (Cisco) not having the protocol enabled on the same network. I do not know allot about spanning tree other than it will not let a network loop bring down the whole network. This is all I would use it for.

[Carsten M]

Hallo !
cisco and hp have the same feature if you use spanning-tree (for the dhcp-problem)
->the option: "fast-spanning-tree" or "spanning-tree-fast" for the user-ports.

you can also use rstp on the hp switch and spanning-tree on the cicso switch at the same time.
the user-ports on the cisco switch must have
"fast-spanning-tree" or "spanning-tree-fast"
and the user-ports on the hp-switch no need a
change from the defaults.

helps

[Thomas Bianco]

!??!

i run a full STP network with 4 DHCP subnets and Zero issues.

Spanning tree protocol determines the route a given packet will take IF THE MAC ADDRESS IS KNOWN. DHCP works fine since it uses a broadcast mac at first, then normal transmission later on.

I can't imagine running a switched network without both DHCP and STP.

[John Anvari]

I read a Cisco documnet that explained spanning tree very well. The issue seems to come into play when the switch is "learning" the port that was just activated. Often a computer is ready before the port is and then the dhcp request times out. I have had this problem more often with Windows 98 systems than win2k or XP.

[cxtwo]

I think Carsten is right here; what you need to do is enable all user ports (ports that do not uplink to other switches) in 'port fast' for Cisco, and 'mode fast' on Procurve.

This allows those ports to go straight into the forwarding state for STP, so they do not go through the usual steps of blocking, learning,etc and come up faster. I guess this would allow the DHCP request to go through without the port changing state.

hth

[Matteo Nasi]

Hi all,

I noticed the same issue with 3 Cisco Catalyst (1 3750, 2 2950) and 1 HP Procurve 2650 configured with multiple vlans and 1 dhcp server that distribute ips to the different subnets.

I read a lot of tech material all around and I think that the portfast option could be a possible solution to the problem, I'm going to test it in the next days.

I'll let you know,

Bye Matteo