How to setup replication between PDC and BDC?

Hi,
The role of PDC and BDC went out with NT4 domains. Active Directory 2000 onwards has what is known as roles. These roles are 5 roles which are unique in the forest/domain.  When you install active directory on another server using Dcpromo, you automatically make that server another domain controller capable of processing user login requests.
However, if the first domain controller was to fail, then the roles that it is hosting can be transferred or seized to any other domain controller in the forest/domain.
Have a read of this article
http://support.microsoft.com/kb/324801
it explains the various roles (FSMO) roles used in active directory

Replication between the two domain controllers automatically happens and is controlled through sites and services.

I set it up just the way you described it. However, the copy of DNCP scope that I created on BDC is not enabled because I cannot have 2 DHCP services running unless I split the range, which my boss does not want to do.

So how will the DHCP part work if PDC goes down? If I am not on site at the moment, how can I ensure that the domain users are are not down with the server?

Thank you!

I would like to setup some automatic roll-over without going through seizing the roles and enabling DHCP manually.
Thanks.

Ok, that is a slightly different question

What you need to do there is to impliment a split scope function.

For this example you create two scopes, one on each server and reserve 50% of the IP addresses for one and 50% for the other. So when clients get an IP address they can get one from either server.

As you mentioned you can't do that because your boss doesn't want split scope. The other way is to restore the copy of the dhcp database onto your other domain conroller ready for a restore if the first one dies.
Or more expensive would be to setup DHCP clustering but that requires two servers and shared storage. I think you may be out of luck on a quick solution.
Also remember that clients can function without a dhcp server running for a short period of time. Depending on your lease time clients connect back to renew their ip address at 50% of the lease time,(as long as they are not rebooted). So you can survive without a DHCP server for a few hours before any noticable problems.
You could setup a robocopy task to backup the dhcp database nightly if you are really keen.
Steven

Also you may need to read up on what roles survive on for how long.
What you are really tallking about is bigger than just dhcp.

you cannot have more than one active DHCP server on a subnet without splitting the scope without potentially creating MASSIVE IP conflict problems.  If you want more than DHCP server split the scope.  Why wouldn't you?  

If Splitting the scope is truly not an option, then you can spend $10,000+ on building a Windows cluster - DHCP is supported in a cluster and then, if one of two servers fails, the other other detects that failure and picks up where the other left off.  (Why $10,000 - you need two copies of Server 2003/2008 Enterprise since clustering is only available in an Enterprise license, plus the hardware to install two copies on - I figure $2,000 is sufficient for a clustered DHCP server... but frankly, this whole idea is just explaining HOW it can be done - I'm not at all recommending it because it's a HORRIBLE cost-to-benefit ratio.

Finally, you should understand how DHCP works.  DHCP leases addresses to devices on the network.  Typical lease times are anywhere between 1 day and 8 days.  DHCP clients attempt to renew the lease after 50% of the lease time has expired.  That means that if you have a 1 day lease and your worker comes in, turns his computer on at 9am, and the server fails at 9:05 am, then the worker will NOT experience a problem UNLESS they reboot their machine (or try to release and renew their IP) until 9am the next morning.

How can I use robocopy to backup DHCP? Will DHCP lease increase help or if the client reboots, it will not matter?
Thank you!

Leew,

Some organizations, especially if using NAP / MAC Authentication / PXE for image distribution require a high available solution for DHCP requiring clustering!  Do not try to eliminate it as an option just because of funding, someone will need to have it HA and will pay to have it so, for some places 10,000 is a drop in the bucket!

I was addressing his question in my perception of his circumstances.  If this person worked in a company where $10,000 was a drop in the bucket, I would expect him to know how DHCP leases worked or he wouldn't be assigned to this project.

If the client reboots, they will have a problem, regardless how long the lease is.  

If a company needs DHCP highly available such a company is ALMOST CERTAINLY going to have at least one cluster to begin with and with DHCP otherwise being a very lightweight service, it could then be easily run off any existing cluster.

So if the environament is stable and there are not many NIC cards changes made, etc, can I increase the DHCP lease for several months, and if the server goes down the users will be able to use the IP addresses for the remainder of time? How about if they reboot? I assume the IP will be the same so that should not be a problem, should it?
THANKS.

etraxler
How many workstations are we talking about?  If you have workstations that need to stay on the network, then just assign them static ip addresses!

Thank you for all of your input.

Does your boss have a good reason not to split the scope? If not, then that would be the most popular and cheapest method to have redundant DHCP services. There is no additional cost involve and you would have redundant services, afteall, you never know when ther user reboot the computer during the outage of your DC w/ DHCP service. Rebooting the PC would release the IP, if no DHCP server available, you would get the automatic private ip addresse and would not have access to other servers.
But one to you could do is create a daily or weekly scheduled task to do the backup of the DHCP database. This way, if you ever need to restore the database of the DHCP to the same or any other DHPC server, all you need to do is import the database and you will have a fully functional DHCP server in just a few minutes.

I wouldn't recommend more than a 10 day lease.  The environment may be stable, but you never know when network changes are going to need to be made - a new gateway, a new DNS server, etc.  As noted before, any client that reboots will have a problem if there is no DHCP servicing clients.  And keep in mind - how long are you thinking the DHCP server would be down?  If I had a client with a DHCP server issue, I'd have that fixed in minutes... or at worst, hours if the server itself failed.  

But if outage happens, then if the scope is split, let's say 50/50, then 50 of the users will not be able to get a new IP, right?

How many network devices do you have and how large is your scope?

If you have 200 devices, a scope of 200 and you split the scope 50/50 and then yes, you have problems - but not, per se because the server failed.  If you reach - I'll go with 125 devices - I would be suggesting that NOW is the time to EITHER (my preference) start subnetting and create a new subnet or extend your subnet to 510 devices by changing your subnet mask from 255.255.255.0 to 255.255.254.  At which point you can assign one server to handle, for example, 192.168.0.50-200 while the other handles 192.168.1.50-200 - since the subnet mask of 255.255.254.0 means these are ALL on the SAME network, you now have scopes of 150 available address on EACH DHCP server a total number of devices that is LESS than 150.

And again, HOW MANY of these devices would end up being rebooted during a minutes to hours outage?

I have about 100 devices on the network. I will have to look at scope range tomorrow. I will get back to you on that one.
Thank you.