Primary DC (FSMO) Fails... what do you do?

[CyBeRkId2002]

Hi all,

 

Currently looking into our disaster recovery plan and, just so I am clear in my head, in the unlikely event our FSMO dies would the following get us up and running with minimum fuss and downtime:

 

• Log on to our second DC, seize FSMO roles and setup DHCP (already has DNS)
  
• Rebuild failed FSMO server, patch, install software, connect up SAN drives, setup shares
  
• Run DCPromo on rebuilt server
  
• Transfer back FSMO (if necessary)
  

 

Not that I would want to do it but it doesn't seem to bad really - days work tops (most of which is patching)... or am I missing bits? We have backups in place etc. using DPM but as we are still running 2003 on our DCs until next summer BMR is out of the question so I am considering what-if scenarios (actually the above probably looks more appealing to me than a BMR, I would worry about replication being out of sync if the PDC was rebuilt).

 

Michael

Edited October 22, 2012 by CyBeRkId2002

[Davit2005]

Rebuilt the PDC Emulator at last place in about 4 hours between 2 of us from a fresh install including rebuilding the RAID. That included restoring all the users data and shares that were on the PDC emulator. We had never done it before and had to find out how to seize the neccesary roles etc and then putting them back on the rebuilt server.

 

Must admit it actually done us a favour as was causing all sorts of grief :-), regulary running out of c drive space and causing replication issues.

Edited October 22, 2012 by Davit2005

[Firefox]

I would also make sure to follow best practice, when seizing the role, make sure the server is rebuilt with a different IP\name. In most cases this shouldn't be necessary but it's advised to have a "clean" install.

 

In some cases you can have a server failure and still be able to move the roles over cleanly. In this case I'd happily keep the same name/ip if the server is rebuilt

[jamesfed]

Or restore from backup??

 

We've done a DR test in the past at a primary we look after with ~300GB of data and a DC restored in a few hours.

[Firefox]

If you only have 1 DC in your environment, then sure a restore from backup is the way to go. However if you have multiple DC's and to the best of your knowledge replication is working. Then I would never try a restore, and always go for a rebuild and repromo.

 

This is unless your backup happened to be minutes before the failure - otherwise you'll have no end of users moaning that their passwords are no longer working, the people who manage your accounts wondering where their days work has gone.

[Dragon]

Do do an non authoritative restore and you would avoid that!

[jamesfed]

Using Ntdsutil.exe to transfer or seize FSMO roles to a domain controller

 

If it is possible, fix the broken domain controller that is assigned the FSMO roles.

[timzim]

Split your DHCP scope 80%/20% across your two DCs. That way if one server goes down the other will still be available to dish out IPs. Saves you from a whole lot of pain if the IPs start expiring but you haven't yet sorted out your DC loss (or if the DC goes down over the weekend).