[EMAIL PROTECTED] wrote:
>
> the dhcp server will check assigned IP's over the network based on the
> global period assigned on the server ( e.g. 3 days). If the IP ( client )
> is still there, it will renew the lease ( for another 3 days, etc.). If
> the client IP is not connected, it will mark that IP address as available,
> and will assign it to another machine as needed. It can also ping an IP
> address BEFORE assigning it. So if you have static IP addresses on the
> net being used by a client, the dhcp server will mark that IP address as
> unusable, and will not assign it to any other machine, to avoid a conflict.
Actually, the CLIENT is entirely responsible for "flapping"(Gulliver's
Travels
reference) the server. At time = lease/2 the client requests a lease
renewal.
If the IP address is available the server will issue a new lease.
Now, with respect to the pinging of ip addresses before issue, the RFC's
(and
possibly MS docs) say that this pinging is supposed to happen (IIRC the
client
is supposed to do the pinging). I have never seen any evidence of this
pinging
happening on our network of win/mac clients and linux/ISC or NT servers.
The
servers will merrily issue in-use IP addresses unless they are
"reserved" in the
dhcpd.conf, and the clients will happily accept them, then turn around
and bitch
about address conflicts.
Now, we mostly use reservations (essentially static, served ip
addresses) so
my experience with dynamically assigned ip is limited to smaller
environments
with 30-160 machines getting dynamic ip addresses. We do run 800-1200
machines
off of static entries in dhcpd.conf in our main production environment.
One thing that would be kind of interesting would be to monitor the dhcp
conversation with tcpdump to see what the difference is between the
linux
and windows requests. Can you even run tcpdump BEFORE you initialize a
NIC?
--
To unsubscribe: mail [EMAIL PROTECTED] with "unsubscribe"
as the Subject.
