Windows leases through a PC reset

[SCOTT BURGIN]

to be clear ( I don't think my original post was after I read it myself ) - the situation is one where the DHCP 
server is up and operational at a corporate headquarters - a remote PC off the corporate WAN is cut off from the headquarters (say a WAN outage has occurred).  What happens what that PC reboots and can't "see" the DHCP server ?

I find Simon's observation that it may check other things (i.e. a ping of its last known default gateway) interesting.
In the case described above, the PC would in fact be able to ping its default gateway it last had (say a local layer 3 switch), but not be able to talk to the DHCP server.  We're going to test this situation - was curious what others may have seen in this situation...

>>> Simon Hobson <dhcp1 at thehobsons.co.uk> 5/8/2007 2:24 PM >>>
Glenn Satchell wrote:

>  >Basic question - RFC aside - does anyone have real world knowledge 
>(personally
>conducted testing preferred) of whether a Windows machine will 
>continue to use a
>valid ISC provided lease through a power down/power up sequence when the DHCP
>server(s) are unreachable ?  Does the version of the Windows OS change the
>result ?



>I just tried this on my test network. Client is Win XP Home SP2. This
>should be very easy to re-create on a test network with a single dhcp
>server and however many clients you want to try.
>
>These are the snooped packets from the PC as it booted. Note the last
>two are from the correct IP address.
>
>OLD-BROADCAST -> BROADCAST    DHCP/BOOTP DHCPREQUEST
>OLD-BROADCAST -> BROADCAST    DHCP/BOOTP DHCPREQUEST
>OLD-BROADCAST -> BROADCAST    DHCP/BOOTP DHCPREQUEST
>dhcp-14-249.uniq.com.au -> BROADCAST    DHCP/BOOTP DHCPINFORM
>dhcp-14-249.uniq.com.au -> BROADCAST    DHCP/BOOTP DHCPINFORM
>
>Running ipconfig /all on the PC after booting shows all the settings
>are still intact. Note that it is possible to configure dhcp to supply
>options that will cause Windows to release the lease when shutdown, but
>this is not the default behaviour in any release.


I've seen this before on the network at my last job - DHCP was down 
for some reason and we didn't notice until the Mac users couldn't do 
certain things.

I also suspect that the clients do more than just "can't reach a DHCP 
server, carry on with last address" - I think they may well check to 
see if the default router is the same (or something similar) to see 
if they are actually still on the same network, and failing that then 
they'll self-assign a link-local address. I think there's an RFC for 
this.



As an aside a Windows PC is VERY sticky with it's address. Even if 
you release the address, it will still include it as the requested 
address in a future discover. It will even request the last static 
assignment when you change from static to dynamic.

I discovered this when diagnosing a problem at a customers site. They 
have a large campus with multiple buildings all linked by gigabit 
ethernet. We had the network management set up a VLAN from the 
customers office to the reception desk in one of the other buildings 
- and the PC would NOT work, and I couldn't figure out why.

My PowerBook got an address straight away which made me even more 
puzzled. Eventually by packet sniffing I figured that their DHCP 
server (Windows) wasn't authoritative and so never NACKed the request 
for a bad address (picked up from the previous network the PC was on) 
that the PC kept asking for, so the PC just never stopped requesting 
it ! In the end I statically configured the PC with a valid address 
for the network, then switched it back to dynamic - hey presto, that 
address now given by DHCP.


This E-Mail has been scanned for viruses.