In-Depth
So You've Had an Operations Master Die...
All is
not lost. Chart your options before you kill the Operations Master for
good.
- By Jeremy Moskowitz
- March 01, 2001
If you're
going to do some scheduled maintenance on a Domain Controller (DC), you
may wish to transfer one of the Single Operations Masters from the box
that's going to be down. You can do this via NTDSUTIL.exe or with the
appropriate MMC snap-in tool. (For more information about seizing or transferring
operations masters, be sure to check out Microsoft
Q255690, "How to View
and Transfer FSMO Roles in the Graphical User Interface," and
Q255504,
"Using Ntdsutil.exe to Seize or Transfer FSMO Roles to a Domain Controller.")
If a domain
controller that houses a Single Operations Master goes belly up, you'll
need to make some judgments about the situation. First, will you be able
to get the machine back up and running?
If you think
you will be able to bring the box back up, your next question should be
how long until it's back in action? If the answer is less than a day or
so, my advice is to just do everything in your power and focus on getting
that machine back up.
If you'll
be unable to get the box back up and running (because, for example, restoring
from tape has failed), then you've pretty much got no choice other than
to have another DC seize ownership. Seizing ownership of a role is, quite
possibly, the most "violent" thing you can do to your Active Directory.
Do so when necessary — but only when absolutely necessary.
Note that
once you seize certain roles, that original box can never be reintroduced
back into the domain. As a good friend likes to say: "Drown it, stir it,
and drown again." That's FDISK to you and me, folks.
Here's a
chart detailing the operations masters, what they do, and when you should
finally bite the bullet, give up and seize a role.
| Operations master |
Description of role |
Domain- or forest-wide role? |
Affects users directly? |
When you should give up and seize the role |
Notes |
| PDC emulator |
Does everything the NT 4.0 PDC did (and more) |
Domain |
Yes |
When your non-Win2K users complain about
the inability to log on |
You can have another DC seize the PDC emulator
role, then transfer back when the original box is back online (if
desired) |
| Schema master |
Makes sure schema changes happen in one place
then get replicated |
Forest |
No |
When you have to make schema modifications
(an infrequent occurrence, one hopes) |
Once role is seized, re-install Win2K Server
on the original DC. Not doing so could corrupt AD |
| Domain naming master |
Makes sure domains get added properly, and
the information is replicated correctly |
Forest |
No |
When you have to add a new domain (an infrequent
occurrence, one hopes) |
1: Always put this role on a DC that's
also a Global Catalog server.
2: Once role is seized, re-install Win2K Server on the original
DC. Not doing so could corrupt AD |
| RID master |
Hands out RIDs (the later portion of a SID)
to objects |
Domain |
No |
You'll know you've run out of RIDs because
you'll see error messages when creating new accounts. This may happen
faster if you have NT 4.0 BDCs in your domain |
Once role is seized, re-install Win2K server
on the original DC. Not doing so could corrupt AD |
| Infrastructure Master |
Tracks the movement of user objects between
OUs and domains |
Domain |
No |
If you get errors moving accounts within
a domain, you'll know it's time. |
1: Don't put this role on a DC that's
also a Global Catalog server.
2: You can have another DC seize the Infrastructure Master
role, then transfer back when the original box is back online (if
desired) |