[cisco-voip] DHCP question

[Mike Newell]

I'm seeing some interesting behavior on our network with our 7970 phones.
The network is implemented on a set of 6509 switches.  The switches have
MSFCs in them that are set up with helper addresses of our DHCP servers.
We have several VLANs that use DHCP services.  The phone ports are tied to
two of those: VLAN-A is where the phones live (the voice VLAN), and VLAN-B
is where the user workstations live (the access VLAN).  The ports are not
set for trunking.

When the phones first come up they apear to issue DHCP queries to the
server from the access VLAN.  A few seconds later they issue DHCPRELEASEs
for their (recently acquired) access VLAN addresses and re-issue
DHCPREQUESTs for the same address from the voice VLAN.  Of course the DHCP
server sees them requesting access VLAN addresses from the voice VLAN and
after mildly chiding them gives them appropriate addresses on the voice
VLAN.

Unfortunately we didn't anticipate this in our DHCP deployment (silly us).
We have another implementation based on 3860 switches which does not
exhibit this behavior.  Our pools were pretty small so we have been
running out of address space 'cause the 50 phones (soon to be >150) keep
camping out address space.

Anyone else seen this?  Anyone know what could be causing it?

There are two L2 differences between the networks: The new one is running
6509s with IOS and access-mode ports, and the older one is running 3860s
with IOS and trunk-mode ports.  I've not tested trunk vs access mode in
this situation (yet).  In VoIP school we were told the phones boot, use
CDP to decide what VLAN to attach to, then do the DHCP dance.  I'm
wondering if what they are REALLY doing is booting, doing the DHCP dance
on the access VLAN, doing the CDP dance, moving to the voice VLAN, then
re-doing the DHCP dance.  If that's the case I sure would like to know if
there's a way to stop that... :-)

Thanks!

Mike