We're in the process of testing for switching DHCP_DNS services over from
QIP to Netware. We have DDNS working, and have also imported some DNS
entries from QIP into the Netware DNS.
We were interested in what would happen if there was an existing DNS entry
for a workstation and we forced the workstation to release and renew its IP
address - would it update the pre-existing DNS entry?
The answer seems to be - no. It occurs to me that the imported *Dynamic*
hostname entries from QIP are appearing in the Netware NS database as
*static* entries, and are thus not being overwritten?
Is this likely to be the case? Are there two types of DNS entry, static and
dynamic?
We are concerned because it is a requirement that workstations here are
registered as hostnames using DDNS because there are a few devices that
require to connect to workstations via hostnames. If we swing over to
Netware DHCP-DNS not all workstations are going to renew immediately, and
they will have no hostname entry. But if we import the QIp DNS database the
workstation hostname entires will be there, but they will be static and
won't update when the workstation renews/changes it's IP address via DHCP.
You see the issue? Am I just misunderstanding how things work and/or is
there a way to do this?
Regards,
Steve Law
If a new workstation gets a DHCP address, a new DNS resource record is
created for it and the IP address is associated with it. For example,
HOSTA is associated with 192.168.1.21. Then, if HOSTA releases it's IP
address, the IP is removed from the DNS RR but the RR is not deleted,
so you're left with HOSTA, but it has no IP address associations.
Later, if HOSTA comes back online, the IP address it's given is
associated with the existing RR. And when it releases that address,
the association is removed, but the DNS RR again remains.
In your situation, what you may need to do is import the old RR's, then
delete IP address associations from them. Then when DHCP gives a new
address to the hosts, I think the RR's will be updated accordingly.
I've not tested this, though, so assuming you do, please let us know
what you find.
bd
NSC Volunteer SysOp
www.InsightNetSolutions.net
Thanks for your response. So what you're saying is: when the workstation
formally releases the IP address, the DNS Hostname remains but the IP
address is stripped from it.
That makes sense. We were actually creating 'false' DNS entries with
invented IP addresses but the same hostname as the workstation. We expected
the workstation, having released and renewed its own IP address, to then
claim the hostname currently associated with the false IP address and
overwrite the false address with its current one. But presumably because no
machine has released that false address it stays in the DNS and won't be
removed or replaced.
Is there a 'clean up' function that removes obsolete host names from DNS? A
timeout for them?
We have a plan to get round the original problem we had, which was: if we
migrate floor-by-floor of our building to Netware DHCP-DNS, then users on
the other floors still using the old QIP servers won't be able to lookup the
hostnames of the workstations moved to the Netware DNS. We believe we can
add in all users workstation DNS settings the new Netware DNS server while
keeping the existing reference to the QIP server. That way devices which
fail to find workstations migrated over to Netware DHCP-DNS in the QIP
server will then check with Netware DNS and find them.
Then once all workstations are registering with Netware DHCP-DNS we can
remove the workstation references to QIP and shut QIP down.
See any problems with that?
Thanks,
Steve Law
"Brad Doster" <b...@NSCSysOps.net> wrote in message
news:VA.000024a...@nscsysops.net...
> But presumably because no
> machine has released that false address it stays in the DNS and won't be
> removed or replaced.
>
Correct, and which makes sense since it is possible for a single host to have
multiple IP addresses.
> Is there a 'clean up' function that removes obsolete host names from DNS? A
> timeout for them?
>
Nope to both.
> See any problems with that?
>
No, your plan sounds perfectly quite workable to me.
"Brad Doster" <b...@NSCSysOps.net> wrote in message
news:VA.000024a...@nscsysops.net...