Need help with DNS design and settings

S

Slater

I've been an NT4 admin for years and am responsible for migrating us to
2003. I'm new to AD and am certainly no expert on AD or DNS, but I have read
MS books, Mark Minasi's 2003 book, and the O'Riely 2003 DNS book. All of the
books and scenerios all seem to just discuss simple AD setups. I had a test
AD working for months using 2000 and it worked great but that was a single
domain. I started over with 2003 this time and now that I am trying to add a
child domain (i.e. domain tree) I am having some problems and no one seems
to cover this scenerio to the point where I understand it. I understand the
CONCEPTS just fine - it's when you get to the nuts and bolts of what
settings go where that everyone seems to leave that out of all of the
training material.

Here's the setup - Each domain will have 2 DCs. The DCs will run Microsoft
DNS and be AD-integrated. One domain (root.priv) will be an empty root
domain for the sole purpose of isolating the Enterprise admin account and
making it easier down the road to reshape the forst if we ever need to. The
second domain (corp.priv) will be the actual production domain that all 100
of my users will belong to and use. It's just a separate tree in the forest.
I also have 2 caching BIND DNS servers on the outside interface for internet
queries, which I will slave my internal DNS servers to for external query
forwarding. Each DC will be a GC server, and one DC in each domain will be
located offsite for disaster recovery purposes (P2P VPN connection between
the sites). That's pretty much it, other than I need to run WINS in the
corp.priv domain. The corp.priv domain will be divided into 2 network
subnets (the second subnet is a QA network that is currently an NT4 domain
but I will just make it an OU in the corp.priv domain once I migrate). So my
plan was to have a DHCP server in the corp.priv domain that will give out
addresses in 2 different subnets (we have DHCP relaying enabled on our cisco
routers). Sounds easy enough on paper, but once I tried to build it I am
having problems with DNS. I've been trying a bunch of things but it's
getting down to crunch time and it's starting to tick me off.

Here's where I'm confused:

1. Local TCP/IP settings on the DNS servers:

I'm confused how to fill out each DC/DNS server's TCP/IP settings. For
example, in the local TCP/IP properties, I know that all DNS servers should
point to themselves as the primary and no secondary, so that's what I've
done on all of the servers (I used the actual IP of the box, not 127.0.0.0
like Microsoft says to do). But I don't know if the same "point to itself as
the primary w/no secondary" rule applies for the corp.priv domain's DNS
servers as well.

2. The DNS settings:

What goes in the name servers box? Do you just list each name server in that
domain, or do you list EVERY name server in your forest in every DNS
server's name server box? For example, on the root.priv DNS servers do I
just list the 2 root.priv servers, and on the corp.priv DNS servers list the
corp.priv servers? Or do I need to instead list all 4 DNS servers on each
DNS server?

What goes in the forwarders box? Since I want to be slaved to external
forwarders for internet queries, I put the address of my 2 external DNS
servers in the forwarders box for "all other DNS domains" and checked the
"do not use recursion for this domain" checkbox. This worked great on the
root.priv DNS servers, but do I do the same on the corp.priv servers?
Basically I want ANY internet query from ANY internal dns server to be
slaved to external forwarders. But I don't understand if child domain DNS
servers are even supposed to resolve internet queries themselves, or if
child domain DNS servers are supposed to forward all DNS queries (internal
or external) to its parent's DNS server?

And how does root.priv and corp.priv forward queries to one another? Someone
recommended to me that I use conditional forwarding. For example, on the
forwarders tab of the root.priv DNS servers, create a new corp.priv domain
and list the corp.priv's DNS servers. And do the opposite for the corp.priv
DNS servers. Is this correct and do I check the "do not use recursion for
this domain" checkbox like I did for the external slave forwarders?

3. Client TCP/IP settings:

What DNS server would clients point to as their primary and secondary DNS
servers? Should machines in Root.test point to the root.test dns servers,
and the machines in corp.test point to the corp.test dns servers? Or should
everyone point to the root.test dns servers? Or does it matter?

=======================

Is there any other tricks I need to do on the 4 DC/DNS servers? A microsoft
article I found said to add the IP addresses (and domain name instead of the
server name) of the DCs to the host file on each DC. This supposedly helps
with DNS resolution issues and some AD replication problems I was having. Is
there any other tips from veterans like this I need to know about to make my
life easier?

Any assistance is greatly appreciated. I know DNS is the heart of AD, and if
that's not worknig then everything else will just be fubarred. So I want to
make sure all of my t's are crossed and i's are dotted before going forward
with any of the NT4 migration procedures and everything else.

I am free to design things however I see fit, so if anyone has a "if I was
going to do it here's what I would do" idea I would love to hear it too.

- Greg
 
M

Mark Renoden [MSFT]

Hi

Just to clarify, we're not talking about parent child domains, we're talking
about forest root and tree root domains. The way you handle these is
different. For your forest root / tree root domain setup ...

1. It's normally good to point the DC/DNS server to itself and to another
DC/DNS server in the same domain as alternate. This way, if the DNS service
fails for any reason locally, you've got the alternate you can make requests
to.

2. The name servers box is usually populated automatically (at least I
thought so) with the DNS servers you have the AD integrated zones on. So
for example, the root.priv AD integrated zone, you should see the two DC/DNS
servers that exist in that domain listed there.

You want to forward to your BIND servers for external resolution in both
domains. To resolve names from one domain to another, consider secondary or
stub zones (this are kewl in 2K3).

3. Clients should point to the DC/DNS servers in their own domain with the
preferred server in the same site.

HTH
--
Mark Renoden [MSFT]
Windows Platform Support Team
Email: (e-mail address removed)

Please note you'll need to strip ".online" from my email address to email
me; I'll post a response back to the group.

This posting is provided "AS IS" with no warranties, and confers no rights.
 
S

Slater

On 1, are you saying the DNS server should point to itself as the primary,
and a second DNS server in the same domain as a secondary (in my example I
have 2 DNS servers per domain, so each one would list the other as the
secondary)? If so, that goes against everything I have ever read. In fact,
when you use the Microsoft wizard, it sets up the DNS server as the primary
with no secondary. I agree with your reasoning, but everyone else must
always say to set it up by itself for a reason...

As far as 2, yes the name servers for each domain do get populated. I am
confused if I should basically list every name server in my entire forest in
the name server box on every DNS server (in my example, there would be 4 DNS
servers total, 2 for each domain. So should I list all 4 servers in the name
server box on all 4 machines?) I was thinking this would help the servers
"find" one another. Or does the name server box list the servers that are
authoritative for that domain only?

Also on #2 what's your opinion on the conditional forwarding I was told to
do? This is an alternative to stub zones I believe and is supposed to help
requests for addresses in another domain easier to locate. For example, if a
user in root.priv asks the root.priv DNS server for the address of a machine
in corp.priv, the root.priv DNS server would forward the request to the
corp.priv DNS server. Correct? Again, the concept sounds logical but I just
wanted to make I was setting everything up right.

I understand #3 - makes sense.

Thanks,
- Slater


Mark Renoden said:
Hi

Just to clarify, we're not talking about parent child domains, we're talking
about forest root and tree root domains. The way you handle these is
different. For your forest root / tree root domain setup ...

1. It's normally good to point the DC/DNS server to itself and to another
DC/DNS server in the same domain as alternate. This way, if the DNS service
fails for any reason locally, you've got the alternate you can make requests
to.

2. The name servers box is usually populated automatically (at least I
thought so) with the DNS servers you have the AD integrated zones on. So
for example, the root.priv AD integrated zone, you should see the two DC/DNS
servers that exist in that domain listed there.

You want to forward to your BIND servers for external resolution in both
domains. To resolve names from one domain to another, consider secondary or
stub zones (this are kewl in 2K3).

3. Clients should point to the DC/DNS servers in their own domain with the
preferred server in the same site.

HTH
--
Mark Renoden [MSFT]
Windows Platform Support Team
Email: (e-mail address removed)

Please note you'll need to strip ".online" from my email address to email
me; I'll post a response back to the group.

This posting is provided "AS IS" with no warranties, and confers no rights.

Slater said:
I've been an NT4 admin for years and am responsible for migrating us to
2003. I'm new to AD and am certainly no expert on AD or DNS, but I have
read
MS books, Mark Minasi's 2003 book, and the O'Riely 2003 DNS book. All of
the
books and scenerios all seem to just discuss simple AD setups. I had a
test
AD working for months using 2000 and it worked great but that was a single
domain. I started over with 2003 this time and now that I am trying to add
a
child domain (i.e. domain tree) I am having some problems and no one seems
to cover this scenerio to the point where I understand it. I understand
the
CONCEPTS just fine - it's when you get to the nuts and bolts of what
settings go where that everyone seems to leave that out of all of the
training material.

Here's the setup - Each domain will have 2 DCs. The DCs will run Microsoft
DNS and be AD-integrated. One domain (root.priv) will be an empty root
domain for the sole purpose of isolating the Enterprise admin account and
making it easier down the road to reshape the forst if we ever need to.
The
second domain (corp.priv) will be the actual production domain that all
100
of my users will belong to and use. It's just a separate tree in the
forest.
I also have 2 caching BIND DNS servers on the outside interface for
internet
queries, which I will slave my internal DNS servers to for external query
forwarding. Each DC will be a GC server, and one DC in each domain will be
located offsite for disaster recovery purposes (P2P VPN connection between
the sites). That's pretty much it, other than I need to run WINS in the
corp.priv domain. The corp.priv domain will be divided into 2 network
subnets (the second subnet is a QA network that is currently an NT4 domain
but I will just make it an OU in the corp.priv domain once I migrate). So
my
plan was to have a DHCP server in the corp.priv domain that will give out
addresses in 2 different subnets (we have DHCP relaying enabled on our
cisco
routers). Sounds easy enough on paper, but once I tried to build it I am
having problems with DNS. I've been trying a bunch of things but it's
getting down to crunch time and it's starting to tick me off.

Here's where I'm confused:

1. Local TCP/IP settings on the DNS servers:

I'm confused how to fill out each DC/DNS server's TCP/IP settings. For
example, in the local TCP/IP properties, I know that all DNS servers
should
point to themselves as the primary and no secondary, so that's what I've
done on all of the servers (I used the actual IP of the box, not 127.0.0.0
like Microsoft says to do). But I don't know if the same "point to itself
as
the primary w/no secondary" rule applies for the corp.priv domain's DNS
servers as well.

2. The DNS settings:

What goes in the name servers box? Do you just list each name server in
that
domain, or do you list EVERY name server in your forest in every DNS
server's name server box? For example, on the root.priv DNS servers do I
just list the 2 root.priv servers, and on the corp.priv DNS servers list
the
corp.priv servers? Or do I need to instead list all 4 DNS servers on each
DNS server?

What goes in the forwarders box? Since I want to be slaved to external
forwarders for internet queries, I put the address of my 2 external DNS
servers in the forwarders box for "all other DNS domains" and checked the
"do not use recursion for this domain" checkbox. This worked great on the
root.priv DNS servers, but do I do the same on the corp.priv servers?
Basically I want ANY internet query from ANY internal dns server to be
slaved to external forwarders. But I don't understand if child domain DNS
servers are even supposed to resolve internet queries themselves, or if
child domain DNS servers are supposed to forward all DNS queries (internal
or external) to its parent's DNS server?

And how does root.priv and corp.priv forward queries to one another?
Someone
recommended to me that I use conditional forwarding. For example, on the
forwarders tab of the root.priv DNS servers, create a new corp.priv domain
and list the corp.priv's DNS servers. And do the opposite for the
corp.priv
DNS servers. Is this correct and do I check the "do not use recursion for
this domain" checkbox like I did for the external slave forwarders?

3. Client TCP/IP settings:

What DNS server would clients point to as their primary and secondary DNS
servers? Should machines in Root.test point to the root.test dns servers,
and the machines in corp.test point to the corp.test dns servers? Or
should
everyone point to the root.test dns servers? Or does it matter?

=======================

Is there any other tricks I need to do on the 4 DC/DNS servers? A
microsoft
article I found said to add the IP addresses (and domain name instead of
the
server name) of the DCs to the host file on each DC. This supposedly helps
with DNS resolution issues and some AD replication problems I was having.
Is
there any other tips from veterans like this I need to know about to make
my
life easier?

Any assistance is greatly appreciated. I know DNS is the heart of AD, and
if
that's not worknig then everything else will just be fubarred. So I want
to
make sure all of my t's are crossed and i's are dotted before going
forward
with any of the NT4 migration procedures and everything else.

I am free to design things however I see fit, so if anyone has a "if I was
going to do it here's what I would do" idea I would love to hear it too.

- Greg
 
M

Mark Renoden [MSFT]

Hi

On point 1, you understand me correctly and it certainly can't hurt to have
the alternate set. The way it works is that the preferred is used until it
doesn't response (note that this is different to a negative response). When
there is no response, we start using the alternate until it doesn't respond.
We then toggle back to the primary.

In troubleshooting issues for customers, we often point all servers at the
same DNS server for the primary and reboot so that they have a consistent
view of DNS. I can't think of any good reason for never using an alternate
provided all DNS servers have a consistent copy of the zone. Under Windows
2000, there was actually an issue with pointing DC's to themselves for DNS
in the forest root domain (islanding). As far as I know, Microsoft has
resolved this in 2003.

2. Name server box is only the servers authoritative for that zone. You'll
notice it's a tab on the properties of the specific zone. Just leave it
alone and you should be OK.

Conditional forwarding will work. I just think it's more administrative
overhead and stub zones are self updating. If you add or remove a DNS
server, the SOA's get updated.

3. Forgot to mention the _msdcs.<forestroot> zone. Under Windows Server
2003, the default is that the zone replicates to all DNS servers in the
forest. This is a good thing. Everyone needs access to this zone.

HTH
--
Mark Renoden [MSFT]
Windows Platform Support Team
Email: (e-mail address removed)

Please note you'll need to strip ".online" from my email address to email
me; I'll post a response back to the group.

This posting is provided "AS IS" with no warranties, and confers no rights.



Slater said:
On 1, are you saying the DNS server should point to itself as the primary,
and a second DNS server in the same domain as a secondary (in my example I
have 2 DNS servers per domain, so each one would list the other as the
secondary)? If so, that goes against everything I have ever read. In fact,
when you use the Microsoft wizard, it sets up the DNS server as the
primary
with no secondary. I agree with your reasoning, but everyone else must
always say to set it up by itself for a reason...

As far as 2, yes the name servers for each domain do get populated. I am
confused if I should basically list every name server in my entire forest
in
the name server box on every DNS server (in my example, there would be 4
DNS
servers total, 2 for each domain. So should I list all 4 servers in the
name
server box on all 4 machines?) I was thinking this would help the servers
"find" one another. Or does the name server box list the servers that are
authoritative for that domain only?

Also on #2 what's your opinion on the conditional forwarding I was told to
do? This is an alternative to stub zones I believe and is supposed to help
requests for addresses in another domain easier to locate. For example, if
a
user in root.priv asks the root.priv DNS server for the address of a
machine
in corp.priv, the root.priv DNS server would forward the request to the
corp.priv DNS server. Correct? Again, the concept sounds logical but I
just
wanted to make I was setting everything up right.

I understand #3 - makes sense.

Thanks,
- Slater


Mark Renoden said:
Hi

Just to clarify, we're not talking about parent child domains, we're talking
about forest root and tree root domains. The way you handle these is
different. For your forest root / tree root domain setup ...

1. It's normally good to point the DC/DNS server to itself and to another
DC/DNS server in the same domain as alternate. This way, if the DNS service
fails for any reason locally, you've got the alternate you can make requests
to.

2. The name servers box is usually populated automatically (at least I
thought so) with the DNS servers you have the AD integrated zones on. So
for example, the root.priv AD integrated zone, you should see the two DC/DNS
servers that exist in that domain listed there.

You want to forward to your BIND servers for external resolution in both
domains. To resolve names from one domain to another, consider secondary or
stub zones (this are kewl in 2K3).

3. Clients should point to the DC/DNS servers in their own domain with
the
preferred server in the same site.

HTH
--
Mark Renoden [MSFT]
Windows Platform Support Team
Email: (e-mail address removed)

Please note you'll need to strip ".online" from my email address to email
me; I'll post a response back to the group.

This posting is provided "AS IS" with no warranties, and confers no rights.

Slater said:
I've been an NT4 admin for years and am responsible for migrating us to
2003. I'm new to AD and am certainly no expert on AD or DNS, but I have
read
MS books, Mark Minasi's 2003 book, and the O'Riely 2003 DNS book. All
of
the
books and scenerios all seem to just discuss simple AD setups. I had a
test
AD working for months using 2000 and it worked great but that was a single
domain. I started over with 2003 this time and now that I am trying to add
a
child domain (i.e. domain tree) I am having some problems and no one seems
to cover this scenerio to the point where I understand it. I understand
the
CONCEPTS just fine - it's when you get to the nuts and bolts of what
settings go where that everyone seems to leave that out of all of the
training material.

Here's the setup - Each domain will have 2 DCs. The DCs will run Microsoft
DNS and be AD-integrated. One domain (root.priv) will be an empty root
domain for the sole purpose of isolating the Enterprise admin account and
making it easier down the road to reshape the forst if we ever need to.
The
second domain (corp.priv) will be the actual production domain that all
100
of my users will belong to and use. It's just a separate tree in the
forest.
I also have 2 caching BIND DNS servers on the outside interface for
internet
queries, which I will slave my internal DNS servers to for external query
forwarding. Each DC will be a GC server, and one DC in each domain will be
located offsite for disaster recovery purposes (P2P VPN connection between
the sites). That's pretty much it, other than I need to run WINS in the
corp.priv domain. The corp.priv domain will be divided into 2 network
subnets (the second subnet is a QA network that is currently an NT4 domain
but I will just make it an OU in the corp.priv domain once I migrate). So
my
plan was to have a DHCP server in the corp.priv domain that will give out
addresses in 2 different subnets (we have DHCP relaying enabled on our
cisco
routers). Sounds easy enough on paper, but once I tried to build it I
am
having problems with DNS. I've been trying a bunch of things but it's
getting down to crunch time and it's starting to tick me off.

Here's where I'm confused:

1. Local TCP/IP settings on the DNS servers:

I'm confused how to fill out each DC/DNS server's TCP/IP settings. For
example, in the local TCP/IP properties, I know that all DNS servers
should
point to themselves as the primary and no secondary, so that's what
I've
done on all of the servers (I used the actual IP of the box, not 127.0.0.0
like Microsoft says to do). But I don't know if the same "point to itself
as
the primary w/no secondary" rule applies for the corp.priv domain's DNS
servers as well.

2. The DNS settings:

What goes in the name servers box? Do you just list each name server in
that
domain, or do you list EVERY name server in your forest in every DNS
server's name server box? For example, on the root.priv DNS servers do
I
just list the 2 root.priv servers, and on the corp.priv DNS servers
list
the
corp.priv servers? Or do I need to instead list all 4 DNS servers on each
DNS server?

What goes in the forwarders box? Since I want to be slaved to external
forwarders for internet queries, I put the address of my 2 external DNS
servers in the forwarders box for "all other DNS domains" and checked the
"do not use recursion for this domain" checkbox. This worked great on the
root.priv DNS servers, but do I do the same on the corp.priv servers?
Basically I want ANY internet query from ANY internal dns server to be
slaved to external forwarders. But I don't understand if child domain DNS
servers are even supposed to resolve internet queries themselves, or if
child domain DNS servers are supposed to forward all DNS queries (internal
or external) to its parent's DNS server?

And how does root.priv and corp.priv forward queries to one another?
Someone
recommended to me that I use conditional forwarding. For example, on
the
forwarders tab of the root.priv DNS servers, create a new corp.priv domain
and list the corp.priv's DNS servers. And do the opposite for the
corp.priv
DNS servers. Is this correct and do I check the "do not use recursion for
this domain" checkbox like I did for the external slave forwarders?

3. Client TCP/IP settings:

What DNS server would clients point to as their primary and secondary DNS
servers? Should machines in Root.test point to the root.test dns servers,
and the machines in corp.test point to the corp.test dns servers? Or
should
everyone point to the root.test dns servers? Or does it matter?

=======================

Is there any other tricks I need to do on the 4 DC/DNS servers? A
microsoft
article I found said to add the IP addresses (and domain name instead
of
the
server name) of the DCs to the host file on each DC. This supposedly helps
with DNS resolution issues and some AD replication problems I was having.
Is
there any other tips from veterans like this I need to know about to make
my
life easier?

Any assistance is greatly appreciated. I know DNS is the heart of AD, and
if
that's not worknig then everything else will just be fubarred. So I
want
to
make sure all of my t's are crossed and i's are dotted before going
forward
with any of the NT4 migration procedures and everything else.

I am free to design things however I see fit, so if anyone has a "if I was
going to do it here's what I would do" idea I would love to hear it
too.

- Greg
 
S

Slater

Re: #1 - OK, that makes sense.

Re: #2, can or should you have stub zones with AD-integrated DNS? I didn't
think you could.

I'm not sure what you mean on #3. I thought the default setting in Server
2003 on ALL zones was to replicate to "All DCs in the domain" (the third
option)?

Finally, a new question - after describing my small network, do you feel I
need reverse lookup zones? Everywhere I read say they are not necessary, but
to me they make sense because a lot of times you resolve names from IPs (I
know I do as the sysadmin). What are the cases when you should run reverse
lookup zones - large companies with dozens of domains and sites?

- Slater


Mark Renoden said:
Hi

On point 1, you understand me correctly and it certainly can't hurt to have
the alternate set. The way it works is that the preferred is used until it
doesn't response (note that this is different to a negative response). When
there is no response, we start using the alternate until it doesn't respond.
We then toggle back to the primary.

In troubleshooting issues for customers, we often point all servers at the
same DNS server for the primary and reboot so that they have a consistent
view of DNS. I can't think of any good reason for never using an alternate
provided all DNS servers have a consistent copy of the zone. Under Windows
2000, there was actually an issue with pointing DC's to themselves for DNS
in the forest root domain (islanding). As far as I know, Microsoft has
resolved this in 2003.

2. Name server box is only the servers authoritative for that zone. You'll
notice it's a tab on the properties of the specific zone. Just leave it
alone and you should be OK.

Conditional forwarding will work. I just think it's more administrative
overhead and stub zones are self updating. If you add or remove a DNS
server, the SOA's get updated.

3. Forgot to mention the _msdcs.<forestroot> zone. Under Windows Server
2003, the default is that the zone replicates to all DNS servers in the
forest. This is a good thing. Everyone needs access to this zone.

HTH
--
Mark Renoden [MSFT]
Windows Platform Support Team
Email: (e-mail address removed)

Please note you'll need to strip ".online" from my email address to email
me; I'll post a response back to the group.

This posting is provided "AS IS" with no warranties, and confers no rights.



Slater said:
On 1, are you saying the DNS server should point to itself as the primary,
and a second DNS server in the same domain as a secondary (in my example I
have 2 DNS servers per domain, so each one would list the other as the
secondary)? If so, that goes against everything I have ever read. In fact,
when you use the Microsoft wizard, it sets up the DNS server as the
primary
with no secondary. I agree with your reasoning, but everyone else must
always say to set it up by itself for a reason...

As far as 2, yes the name servers for each domain do get populated. I am
confused if I should basically list every name server in my entire forest
in
the name server box on every DNS server (in my example, there would be 4
DNS
servers total, 2 for each domain. So should I list all 4 servers in the
name
server box on all 4 machines?) I was thinking this would help the servers
"find" one another. Or does the name server box list the servers that are
authoritative for that domain only?

Also on #2 what's your opinion on the conditional forwarding I was told to
do? This is an alternative to stub zones I believe and is supposed to help
requests for addresses in another domain easier to locate. For example, if
a
user in root.priv asks the root.priv DNS server for the address of a
machine
in corp.priv, the root.priv DNS server would forward the request to the
corp.priv DNS server. Correct? Again, the concept sounds logical but I
just
wanted to make I was setting everything up right.

I understand #3 - makes sense.

Thanks,
- Slater


Mark Renoden said:
Hi

Just to clarify, we're not talking about parent child domains, we're talking
about forest root and tree root domains. The way you handle these is
different. For your forest root / tree root domain setup ...

1. It's normally good to point the DC/DNS server to itself and to another
DC/DNS server in the same domain as alternate. This way, if the DNS service
fails for any reason locally, you've got the alternate you can make requests
to.

2. The name servers box is usually populated automatically (at least I
thought so) with the DNS servers you have the AD integrated zones on. So
for example, the root.priv AD integrated zone, you should see the two DC/DNS
servers that exist in that domain listed there.

You want to forward to your BIND servers for external resolution in both
domains. To resolve names from one domain to another, consider
secondary
or
stub zones (this are kewl in 2K3).

3. Clients should point to the DC/DNS servers in their own domain with
the
preferred server in the same site.

HTH
--
Mark Renoden [MSFT]
Windows Platform Support Team
Email: (e-mail address removed)

Please note you'll need to strip ".online" from my email address to email
me; I'll post a response back to the group.

This posting is provided "AS IS" with no warranties, and confers no rights.

I've been an NT4 admin for years and am responsible for migrating us to
2003. I'm new to AD and am certainly no expert on AD or DNS, but I have
read
MS books, Mark Minasi's 2003 book, and the O'Riely 2003 DNS book. All
of
the
books and scenerios all seem to just discuss simple AD setups. I had a
test
AD working for months using 2000 and it worked great but that was a single
domain. I started over with 2003 this time and now that I am trying
to
add
a
child domain (i.e. domain tree) I am having some problems and no one seems
to cover this scenerio to the point where I understand it. I understand
the
CONCEPTS just fine - it's when you get to the nuts and bolts of what
settings go where that everyone seems to leave that out of all of the
training material.

Here's the setup - Each domain will have 2 DCs. The DCs will run Microsoft
DNS and be AD-integrated. One domain (root.priv) will be an empty root
domain for the sole purpose of isolating the Enterprise admin account and
making it easier down the road to reshape the forst if we ever need to.
The
second domain (corp.priv) will be the actual production domain that all
100
of my users will belong to and use. It's just a separate tree in the
forest.
I also have 2 caching BIND DNS servers on the outside interface for
internet
queries, which I will slave my internal DNS servers to for external query
forwarding. Each DC will be a GC server, and one DC in each domain
will
be
located offsite for disaster recovery purposes (P2P VPN connection between
the sites). That's pretty much it, other than I need to run WINS in the
corp.priv domain. The corp.priv domain will be divided into 2 network
subnets (the second subnet is a QA network that is currently an NT4 domain
but I will just make it an OU in the corp.priv domain once I
migrate).
So
my
plan was to have a DHCP server in the corp.priv domain that will give out
addresses in 2 different subnets (we have DHCP relaying enabled on our
cisco
routers). Sounds easy enough on paper, but once I tried to build it I
am
having problems with DNS. I've been trying a bunch of things but it's
getting down to crunch time and it's starting to tick me off.

Here's where I'm confused:

1. Local TCP/IP settings on the DNS servers:

I'm confused how to fill out each DC/DNS server's TCP/IP settings. For
example, in the local TCP/IP properties, I know that all DNS servers
should
point to themselves as the primary and no secondary, so that's what
I've
done on all of the servers (I used the actual IP of the box, not 127.0.0.0
like Microsoft says to do). But I don't know if the same "point to itself
as
the primary w/no secondary" rule applies for the corp.priv domain's DNS
servers as well.

2. The DNS settings:

What goes in the name servers box? Do you just list each name server in
that
domain, or do you list EVERY name server in your forest in every DNS
server's name server box? For example, on the root.priv DNS servers do
I
just list the 2 root.priv servers, and on the corp.priv DNS servers
list
the
corp.priv servers? Or do I need to instead list all 4 DNS servers on each
DNS server?

What goes in the forwarders box? Since I want to be slaved to external
forwarders for internet queries, I put the address of my 2 external DNS
servers in the forwarders box for "all other DNS domains" and checked the
"do not use recursion for this domain" checkbox. This worked great on the
root.priv DNS servers, but do I do the same on the corp.priv servers?
Basically I want ANY internet query from ANY internal dns server to be
slaved to external forwarders. But I don't understand if child domain DNS
servers are even supposed to resolve internet queries themselves, or if
child domain DNS servers are supposed to forward all DNS queries (internal
or external) to its parent's DNS server?

And how does root.priv and corp.priv forward queries to one another?
Someone
recommended to me that I use conditional forwarding. For example, on
the
forwarders tab of the root.priv DNS servers, create a new corp.priv domain
and list the corp.priv's DNS servers. And do the opposite for the
corp.priv
DNS servers. Is this correct and do I check the "do not use recursion for
this domain" checkbox like I did for the external slave forwarders?

3. Client TCP/IP settings:

What DNS server would clients point to as their primary and secondary DNS
servers? Should machines in Root.test point to the root.test dns servers,
and the machines in corp.test point to the corp.test dns servers? Or
should
everyone point to the root.test dns servers? Or does it matter?

=======================

Is there any other tricks I need to do on the 4 DC/DNS servers? A
microsoft
article I found said to add the IP addresses (and domain name instead
of
the
server name) of the DCs to the host file on each DC. This supposedly helps
with DNS resolution issues and some AD replication problems I was having.
Is
there any other tips from veterans like this I need to know about to make
my
life easier?

Any assistance is greatly appreciated. I know DNS is the heart of AD, and
if
that's not worknig then everything else will just be fubarred. So I
want
to
make sure all of my t's are crossed and i's are dotted before going
forward
with any of the NT4 migration procedures and everything else.

I am free to design things however I see fit, so if anyone has a "if
I
was
going to do it here's what I would do" idea I would love to hear it
too.

- Greg
 
K

Kevin D. Goodknecht Sr. [MVP]

In
Slater said:
Re: #1 - OK, that makes sense.

Re: #2, can or should you have stub zones with
AD-integrated DNS? I didn't think you could.

I'm not sure what you mean on #3. I thought the default
setting in Server 2003 on ALL zones was to replicate to
"All DCs in the domain" (the third option)?

Finally, a new question - after describing my small
network, do you feel I need reverse lookup zones?
Everywhere I read say they are not necessary, but to me
they make sense because a lot of times you resolve names
from IPs (I know I do as the sysadmin). What are the
cases when you should run reverse lookup zones - large
companies with dozens of domains and sites?

Reverse zones are not required for proper AD function, but without a reverse
zone, your machines will attempt to register in prisoner.iana.org (for the
private addresses) and Win2k3 and XP will log errors 40960 and 40961 in the
event log. If these machines have a public address they will attempt to
register PTRs in the Authoritative DNS for the reverse lookup of the IP
address (usually your ISP's DNS) and log these events.
 
M

Mark Renoden [MSFT]

Hi

2. Stubs can be whatever you like. AD Integrated vs not AD Integrated is
just a question of where it's stored and you can use this to define the
scope of servers ... all DC's or all DNS servers etc.

3. AFAIK the default replication scope is to all DNS server in the forest
for the _msdcs.<forestroot> zone. It needs to be visible to all forest
members so this makes sense. All DNS servers in the forest have a copy and
then all clients that query their local domain DNS server can see it.

I think your reverse lookup zone question was answered.

Kind regards
--
Mark Renoden [MSFT]
Windows Platform Support Team
Email: (e-mail address removed)

Please note you'll need to strip ".online" from my email address to email
me; I'll post a response back to the group.

This posting is provided "AS IS" with no warranties, and confers no rights.

Slater said:
Re: #1 - OK, that makes sense.

Re: #2, can or should you have stub zones with AD-integrated DNS? I didn't
think you could.

I'm not sure what you mean on #3. I thought the default setting in Server
2003 on ALL zones was to replicate to "All DCs in the domain" (the third
option)?

Finally, a new question - after describing my small network, do you feel I
need reverse lookup zones? Everywhere I read say they are not necessary,
but
to me they make sense because a lot of times you resolve names from IPs (I
know I do as the sysadmin). What are the cases when you should run reverse
lookup zones - large companies with dozens of domains and sites?

- Slater


Mark Renoden said:
Hi

On point 1, you understand me correctly and it certainly can't hurt to have
the alternate set. The way it works is that the preferred is used until it
doesn't response (note that this is different to a negative response). When
there is no response, we start using the alternate until it doesn't respond.
We then toggle back to the primary.

In troubleshooting issues for customers, we often point all servers at
the
same DNS server for the primary and reboot so that they have a consistent
view of DNS. I can't think of any good reason for never using an alternate
provided all DNS servers have a consistent copy of the zone. Under Windows
2000, there was actually an issue with pointing DC's to themselves for
DNS
in the forest root domain (islanding). As far as I know, Microsoft has
resolved this in 2003.

2. Name server box is only the servers authoritative for that zone. You'll
notice it's a tab on the properties of the specific zone. Just leave it
alone and you should be OK.

Conditional forwarding will work. I just think it's more administrative
overhead and stub zones are self updating. If you add or remove a DNS
server, the SOA's get updated.

3. Forgot to mention the _msdcs.<forestroot> zone. Under Windows Server
2003, the default is that the zone replicates to all DNS servers in the
forest. This is a good thing. Everyone needs access to this zone.

HTH
--
Mark Renoden [MSFT]
Windows Platform Support Team
Email: (e-mail address removed)

Please note you'll need to strip ".online" from my email address to email
me; I'll post a response back to the group.

This posting is provided "AS IS" with no warranties, and confers no rights.



Slater said:
On 1, are you saying the DNS server should point to itself as the primary,
and a second DNS server in the same domain as a secondary (in my
example I
have 2 DNS servers per domain, so each one would list the other as the
secondary)? If so, that goes against everything I have ever read. In fact,
when you use the Microsoft wizard, it sets up the DNS server as the
primary
with no secondary. I agree with your reasoning, but everyone else must
always say to set it up by itself for a reason...

As far as 2, yes the name servers for each domain do get populated. I
am
confused if I should basically list every name server in my entire forest
in
the name server box on every DNS server (in my example, there would be
4
DNS
servers total, 2 for each domain. So should I list all 4 servers in the
name
server box on all 4 machines?) I was thinking this would help the servers
"find" one another. Or does the name server box list the servers that are
authoritative for that domain only?

Also on #2 what's your opinion on the conditional forwarding I was told to
do? This is an alternative to stub zones I believe and is supposed to help
requests for addresses in another domain easier to locate. For example, if
a
user in root.priv asks the root.priv DNS server for the address of a
machine
in corp.priv, the root.priv DNS server would forward the request to the
corp.priv DNS server. Correct? Again, the concept sounds logical but I
just
wanted to make I was setting everything up right.

I understand #3 - makes sense.

Thanks,
- Slater


Hi

Just to clarify, we're not talking about parent child domains, we're
talking
about forest root and tree root domains. The way you handle these is
different. For your forest root / tree root domain setup ...

1. It's normally good to point the DC/DNS server to itself and to another
DC/DNS server in the same domain as alternate. This way, if the DNS
service
fails for any reason locally, you've got the alternate you can make
requests
to.

2. The name servers box is usually populated automatically (at least I
thought so) with the DNS servers you have the AD integrated zones on. So
for example, the root.priv AD integrated zone, you should see the two
DC/DNS
servers that exist in that domain listed there.

You want to forward to your BIND servers for external resolution in both
domains. To resolve names from one domain to another, consider secondary
or
stub zones (this are kewl in 2K3).

3. Clients should point to the DC/DNS servers in their own domain with
the
preferred server in the same site.

HTH
--
Mark Renoden [MSFT]
Windows Platform Support Team
Email: (e-mail address removed)

Please note you'll need to strip ".online" from my email address to email
me; I'll post a response back to the group.

This posting is provided "AS IS" with no warranties, and confers no
rights.

I've been an NT4 admin for years and am responsible for migrating us to
2003. I'm new to AD and am certainly no expert on AD or DNS, but I have
read
MS books, Mark Minasi's 2003 book, and the O'Riely 2003 DNS book.
All
of
the
books and scenerios all seem to just discuss simple AD setups. I had a
test
AD working for months using 2000 and it worked great but that was a
single
domain. I started over with 2003 this time and now that I am trying to
add
a
child domain (i.e. domain tree) I am having some problems and no one
seems
to cover this scenerio to the point where I understand it. I understand
the
CONCEPTS just fine - it's when you get to the nuts and bolts of what
settings go where that everyone seems to leave that out of all of
the
training material.

Here's the setup - Each domain will have 2 DCs. The DCs will run
Microsoft
DNS and be AD-integrated. One domain (root.priv) will be an empty root
domain for the sole purpose of isolating the Enterprise admin
account
and
making it easier down the road to reshape the forst if we ever need to.
The
second domain (corp.priv) will be the actual production domain that all
100
of my users will belong to and use. It's just a separate tree in the
forest.
I also have 2 caching BIND DNS servers on the outside interface for
internet
queries, which I will slave my internal DNS servers to for external
query
forwarding. Each DC will be a GC server, and one DC in each domain will
be
located offsite for disaster recovery purposes (P2P VPN connection
between
the sites). That's pretty much it, other than I need to run WINS in the
corp.priv domain. The corp.priv domain will be divided into 2
network
subnets (the second subnet is a QA network that is currently an NT4
domain
but I will just make it an OU in the corp.priv domain once I migrate).
So
my
plan was to have a DHCP server in the corp.priv domain that will
give
out
addresses in 2 different subnets (we have DHCP relaying enabled on our
cisco
routers). Sounds easy enough on paper, but once I tried to build it
I
am
having problems with DNS. I've been trying a bunch of things but
it's
getting down to crunch time and it's starting to tick me off.

Here's where I'm confused:

1. Local TCP/IP settings on the DNS servers:

I'm confused how to fill out each DC/DNS server's TCP/IP settings. For
example, in the local TCP/IP properties, I know that all DNS servers
should
point to themselves as the primary and no secondary, so that's what
I've
done on all of the servers (I used the actual IP of the box, not
127.0.0.0
like Microsoft says to do). But I don't know if the same "point to
itself
as
the primary w/no secondary" rule applies for the corp.priv domain's DNS
servers as well.

2. The DNS settings:

What goes in the name servers box? Do you just list each name server in
that
domain, or do you list EVERY name server in your forest in every DNS
server's name server box? For example, on the root.priv DNS servers do
I
just list the 2 root.priv servers, and on the corp.priv DNS servers
list
the
corp.priv servers? Or do I need to instead list all 4 DNS servers on
each
DNS server?

What goes in the forwarders box? Since I want to be slaved to external
forwarders for internet queries, I put the address of my 2 external DNS
servers in the forwarders box for "all other DNS domains" and
checked
the
"do not use recursion for this domain" checkbox. This worked great
on
the
root.priv DNS servers, but do I do the same on the corp.priv
servers?
Basically I want ANY internet query from ANY internal dns server to be
slaved to external forwarders. But I don't understand if child
domain
DNS
servers are even supposed to resolve internet queries themselves, or if
child domain DNS servers are supposed to forward all DNS queries
(internal
or external) to its parent's DNS server?

And how does root.priv and corp.priv forward queries to one another?
Someone
recommended to me that I use conditional forwarding. For example, on
the
forwarders tab of the root.priv DNS servers, create a new corp.priv
domain
and list the corp.priv's DNS servers. And do the opposite for the
corp.priv
DNS servers. Is this correct and do I check the "do not use
recursion
for
this domain" checkbox like I did for the external slave forwarders?

3. Client TCP/IP settings:

What DNS server would clients point to as their primary and
secondary
DNS
servers? Should machines in Root.test point to the root.test dns
servers,
and the machines in corp.test point to the corp.test dns servers? Or
should
everyone point to the root.test dns servers? Or does it matter?

=======================

Is there any other tricks I need to do on the 4 DC/DNS servers? A
microsoft
article I found said to add the IP addresses (and domain name
instead
of
the
server name) of the DCs to the host file on each DC. This supposedly
helps
with DNS resolution issues and some AD replication problems I was
having.
Is
there any other tips from veterans like this I need to know about to
make
my
life easier?

Any assistance is greatly appreciated. I know DNS is the heart of
AD,
and
if
that's not worknig then everything else will just be fubarred. So I
want
to
make sure all of my t's are crossed and i's are dotted before going
forward
with any of the NT4 migration procedures and everything else.

I am free to design things however I see fit, so if anyone has a "if I
was
going to do it here's what I would do" idea I would love to hear it
too.

- Greg
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Top