Restoring all domain controllers in a Windows domain (on vmware)

I feel like there needs to be a MUCH more clear way of setting this up and preforming it. From Veeam and Microsoft both.   I’m talking fail safe step by step hold my hand instructions.

I bet a lot of people are not set up for successes in the event they need to restore their DC’s. 

Interesting thread, We have the issue that after starting our DR site (which is built using a Veeam replication job from backup copies), the sysvol share is not replicating anymore (we have 3 DC’s). I can only get it to replicate again by adding the following regkeys to the DC which holds the FSMO roles:

reg add "HKLM\System\CurrentControlSet\Services\DFSR\Restore" /v SYSVOL /t REG_SZ /d "authoritative"
reg add "HKLM\SYSTEM\CurrentControlSet\Control\BackupRestore\SystemStateRestore" /v LastRestoreId /t REG_SZ /d "10000000-0000-0000-0000-000000000000" 

After a reboot of the DC I added these keys to sysvol is replicating again.

Is this also a known issue?

@MckITGuys thanks for the feedback.
Your case is a classic one for an AD architecture.Usually when a single Domain Controller accuses a problem and you cannot restore all functionality properly, it is easier to perform a "force removal" of the corrupted DC.

Force removal guide:

- Force removal Computer Object DC from ADUC
-DELETE Computer Object

https://techcommunity.microsoft.com/t5/itops-talk-blog/step-by-step-manually-removing-a-domain-controller-server/ba-p/280564 
Perform metadata cleanup

Proceed with the following checks:

- From the DNS console check & delete all entries pertaining to the removed Domain Controller.
- Edit ADSI Edit: Configuration Partition and verify that the DC object has been properly removed
Delete entry from Name Server tab
Delete the SRV records :
DNS: _msdcs.domainname.com -> CNAME (GUID - fqdn) of the DC
DNS: _msdcs.domainname.com -> dc -> _sites -> sitename -> _tcp -> SRV Records for Kerberos and LDAP
DNS: _msdcs.domainname.com -> domains -> domain guid -> _tcp -> LDAP SRV Records
DNS: _msdcs.domainname.com -> gc -> _sites and _tcp (if the DC was also GC)
DNS: _msdcs.domainname.com -> pdc -> _tcp -> LDAP SRV (if the DC was also PDC)
DNS: domainname.com -> A Record for IPv4 and IPv6

- Open Active Directory Site & Service delete in the afferent site the DC server object from the replication partners (this object is not automatically deleted.

- Verify that all AD post force removal features work correctly from replicas to FSMO role and AD time sync.

I attach a few commands for verification:

netdom query fsmo
nltest /dclist:yourdomain.local
nltest /dsgetdc:mydomain.local /force /gc

repadmin /showrepl
repadmin /replsummary
repadmin /replsummary
repadmin /Queue

How to Force Active Directory Replication.

This will do a pull replication, which means it will pull updates from DC2 to DC1.
REPADMIN
REPADMIN
If you want to push replication you will use the /P switch. For example if you make changes on DC1 and want to replicate those to other DCs use this command.

repadmin /syncall dc1 /APeD

repadmin /replsummary > c:\it\replsummary.txt
repadmin /istg * /verbose
DCDIAG

dcdiag /test:dns                    
dcdiag /test:netlogons
dcdiag /test:advertising /v
dfsrdiag ReplicationState /member:yourDCname 

dfsrdiag ReplicationState /member:yourDCname 
dcdiag /test:topology
dcdiag /test:replications
dcdiag /c /e /v >> c:\ADHealth.txt

DNSLint is a Microsoft Windows utility that helps you to diagnose common DNS name resolution issues.
    dnslint /ad /s #IPAddressOfServer#

Proceed to run a new DC promo to replace the DC decom
Check post DCPRMO

repadmin /showrepl /repsto

Get-WinEvent -LogName "Directory Service" | ?{$_.Id -eq 1119} | FL

Get-WinEvent -LogName "DFS Replication" | ?{$_.Id -eq 4604} | FL

dcdiag /c /d /v

DCDiag (part of WS03 SP1 Support tools) displays all information about Domain Controller information.
dcdiag.exe /V /C /D /E /s:#DomainControllerName# > c:\dcdiag.log

NetDiag provides information about specific network configuration for the local machine.
netdiag.exe /v > c:\netnetdiag.log

RepAdmin helps diagnise AD replication issues with WS03 and WS08 DC's.
repadmin.exe /showrepl dc* /verbose /all /intersite > c:\repl.txt
repadmin /syncall -APed.
 
Regards

Hi all,

First off, big thanks to everyone who jumped in with advice.  Learned a lot through this exercise (not exactly my kind of exercise though :-)  I need to read back through the thread and mark a few more replies but since am short on time this morning, will just give a summary of what happened.  Note that the conclusion did NOT need to use Veeam for restores but in case someone hits this thread and wants to know the outcome, it might help them.  As well a caveat - this might not have been the “best” way to do this (i.e. manually pruning AD) but it was the first step we opted to take instead of getting into restoring DCs.

Reminder: we had one bad DC in our environment that we could not log into no-how so we could not demote it gracefully.  This happened about 3 weeks ago and we were starting to experience replication problems (changes not being replicated) so we really needed to get it fixed.  Steps taken:

-shut down the bad DC; before this fix, doing so would cause a significant problem on one of our servers that was using LDAP to this server for authentication (could not get around this) even though this server was *hardcoded* for it’s LDAP settings (could not even get support for this product to figure out why).

-go into ADUC and delete the bad DC from the Domain Controllers OU

-go into DNS and see if this DC was removed from DNS: only one instance was removed correctly (at the root) but all lower instances of it in all kinds of nodes were not removed; manually go through the forward and reverse trees and manually delete all references to the old DC; note that some might think this is extreme but we were going to permanently get rid of this DC anyhow (when we build back another DC, it will have a different name)

-go into Sites and Services (SaS) and remove the bad DC from the pertinent site; remove any connection objects between the bad DC and other DCs

-at this point I stopped and fired up the server that would always “choke” because it was looking for LDAP services from the bad server; this time, it worked fine without it - it reverted back to the other DC at that site

-went into SaS again and fired the replication connections between the remaining DCs; at first I got an error message but that eventually went away - not sure if somehow replication of the context object finally happened enough to allow full replication to happen and I just needed to wait, or did all my poking about finally get it going.  I really did not know if I had to somehow fix things up for the sites or not - but finally it worked (if anyone has any insight into that, I would not mind knowing).

-at this point, it looked as though replication was working correctly via issuing repadmin /replsummary and dcdiag etc.

-all servers stopped having errors and nslookups and all tests seemed okay; newly created objects on the one DC finally showed up on the other DCs (and obviously changes as well);

-one interesting item was that just before the fix took place last week, a few users including myself, had our mapped network drives disappear to one of our servers and we could not re-create them at all; so somehow some trust relationship or something was starting to break down such that our mapped drives stopped working (if anyone has insight on this as well, would like to hear)

I still have a few dcdiag warnings to look at this morning - but when reading a bit on each, none seemed serious (more warnings for best practices etc.); I might post those as another post if they are related to this situation.

Thanks a lot everyone for the help.  I learned a lot more about what I might have needed to do for a Veeam restore of all DCs - glad I did not have to do it.  I also feel like I need to summarize some of the thread items here to have handy in case we need it.  Hopefully I can get to that before this gets too stale in my mind.

Albert (for McKITGuys)

Wondering how it worked out for you. 

Don’t restore from a snapshot? So don’t use Veaam at all essentially is what you are saying?   I won’t be powering down my DC’s to backup daily jobs. 

Yes, I realize powering all of them down and doing a snapshot is ideal, but not realistic in most 24/7 environments. The same reason we have multiple DC’s.

I found the ideal solution is to give one DC the FSMO roles (unless your environment is HUGE and you want to split them up. but under a few thousand users this is probably fine.)  Power off ALL of the DC’s, and restore that specific VM as authoritative.  The Veaam docs \support worked 100% for us in a crit sit with everyone down. 

Hi

VMs, I have restore full AD a couple of times. No issues, particularly if you are using Windows 2016 and above.

The “new” features VM-Generation will control the changes in the AD when is to restore.

When this ID change?

Scenario VM-Generation – ID Change

• VMware vMotion®/VMware vSphere Storage vMotion / Hyper-V Live Migration – No
• Virtual machine pause/resume – No
• Virtual machine reboot – No
• vSphere host reboot – No
• Delete VM Snapshot – No
• Import virtual machine – Yes
• Cold clone – Yes
• Hot clone – Yes
  Note: Either Microsoft or VMware do not support virtual domain controllers hot cloning. Do not attempt hot cloning under any circumstances.
• New virtual machine from VMware Virtual Disk Development Kit (VMDK) copy – Yes
• Cold snapshot revert (while powered off or while running and not taking a memory snapshot) – Yes
• Hot snapshot revert (while powered on with a memory snapshot) – Yes
• Restore from virtual machine level backup – Yes
• Virtual machine replication (using both host-based and array-level replication) – Yes

   

Before we restore the broken Domain Controller, some Domain Controllers restore Best Practices.

• If you are restoring a broken DC in a multi DCs environment, then you can do a normal restore and follow then next restore process.
• If you are restoring all DCs in your Domain, then you should first restore the one with all RMSO roles (this information you should have before you lost your DCs).
• If you are restoring the DC that had the FMSO roles and you want to keep that way(meaning all the existing DCs will sync from your restored DC), then you need to do follow Microsoft How to recover authoritative restore .

So when you need to restore domains controllers, always try to restore the last one that was backup.

I have written a couple of articles regarding this procedure.

Wow. Quite a comprehensive thread this has gotten. Very interesting topic.

To me the most important part of the question to me in the first place was:

All (!) DCs have to be restored at once.

If you try that with VBR without further measures it will fail.

Not a single one of the DCs will bring up the domain. This is because VBR - even since V2 - is smart enough to always do a non-authoritative-restore. This will bring up the DC in a state which first wants to copy the full domain from another DC still alive. You can take note of the process by the additional reboot a DC always does after being recovered by VBR.

Without this full sync of the domain DB your recovered DC will be a useless brick.

Problem is: If you recovered all of them, they will all look for a surviving DC - but there is none.

You then have to put one of them into an authoritative restore manually afterwards by e.g. setting the “Burflags”. Have that process in your emergency guideline just in case...

Veeam btw CAN do an authoritative restore. You can opt to do that in a SureBackup application group. Otherwise every SureBackup test would also come up with useless DCs. Keep in mind that if you have more than one DC in a SureBackup application group, only one should be brought up authoritative while the others must be set to non-authoritative. Then those will copy the domain from the authoritative one.

SureBackup can even be used to automate the authoritative restore of a DC. You might just move the DC from the lab into the “real” world AKA network. Do it carefully. No other DC must be alive. Otherwise your domain will break be getting inconsistent. 

Have a secondary DNS without AD integration for your Veeam components for that. Otherwise without DNS, VBR and SureBackup will not function as expected… We always put a secondary DNS on top of the VBR server to be used only as a backup for the VI and Veeam. It’s just a service...

I am writing up a “playbook” now with a couple different scenarios so forgive me for random questions that may or may not be the “best plan” i.e. I just want to know more in case I have to go down different routes:

-if I keep one DC (the one with most of the newest changes), and just down and delete the other 2 DCs, I am going to have DNS entries for the other two DCs still in DNS (I assume that because the other 2 DCs are at a different site and even if I demote them and remove them from the domain, it is possible that their records will persist on the good DC

-question from that: other than the A record in the root of the zone, do I have to find and remove every reference for each of the downed DCs (service records etc.) or is there one spot to delete it such that it cascade deletes to all service records?

AD should keep a copy of all across all servers.  If there are broken trusts, that should be replicated throughout to all DC’s in AD, so it should not matter which DC is brought online unless we’re talking about changes being made in a narrow window before they are replicated to the other DC’s.  The aside would be if replication had already failed and different DC’s had different copies of AD that were not being replicated amongst themselves.  As previously noted, if this has happened for too long, then the DC’s will have reached their tombstone lifetime and really are of little to no use anyway.  Seizing FSMO roles is trivial, as is cleaning up AD from other DC’s that are no longer available.  Taking down all of the DC’s no longer desired and restoring the most desirable DC authoratively would likely be the way to go.  Seize FSMO, cleanup AD from the unwanted DC’s to prevent any further chances of replication, and then proceed with building new DC’s.

This seems unusual for sure.  But that said, if you have a failed/corrupted DC, power it off and blow it away, clean that DC out of AD and build a new one.  I’d typically suggest using a new name to prevent an possible rogue data from infiltrating back into things….probably not likely, but you’re already in a precarious position, best to not add any possible complications.

You plan looks very similar to what I would do.  But I don’t think I’d try and delete AD data from existing DC’s and reuse them.  I’d just build new.  I suppose you could forcibly remove AD from a DC and reuse it, by why take the risk?  As previously noted, they’re practically throwaway anyway.

Obviously, I can’t make a good call without knowing details on the number of DC’s you have, how many are local vs remote sites, how many remote sites, the root cause of the problem, etc.  And clearly we’ve gone well beyond the scope of restoring DC’s in Veeam.  If you know the root cause of the problem and have resolved it, and you know that AD data is valid as of a certain date, and that you know that the remote site DC’s can either be restored non-authoratively and sync with the DC that is authoratively restored, then I suppose you could resuse them.  But I’d be hesitant to do so in a more complex environment where changes of a rogue DC could cause wrinkles and you have to start the process over again.  I don’t know enough of your situation to make a call, but it sounds more complex than I’d want to take the risk for, and would rather take a safe route and restore the most likely good DC, validate it (offline if I must), and then blow away the others and join them to this one.

Yeah, I shudder at the thought of restoring all DC’s.  I’d honestly consider shutting down all of the DC’s, restoring one authoratively, and then cleaning up AD and building new DC’s for any remote systems.  Building a DC is practically throw-away anymore, especially since you don’t have to do metadata cleanups anymore.

https://www.dell.com/support/kbdoc/it-it/000136611/resetting-the-directory-services-restore-mode-administrator-password

More questions from excellent discussion above:

-netdom resetpwd - so I have a server or user PC that has lost its trust relationship: am I going to be able to log into that server to perform this command?  and if I do, say my server is named DC1 and my domain admin account is SkinnyAdmin, would the command be:

netdom resetpwd /s:DC1 /ud:mydomain.com\SkinnyAdmin /pd:bigFatPassword

-so I assume this from what I have read, since this is a machine password, gets a new generated password from the DC and updates that locally and also in that computer’s object in AD? just checking

-and I assume that this resets the trust relationship at the same time

DSRM password: I have taken over admin of a client’s  network: although I have the administrator passwords, I have not found any DSRM passwords recorded anywhere; I can guess by a list of commonly used passwords but cannot be sure; is that going to prevent a Veeam restore or does it just use the domain admin passwords stored in its credentials setup?  I know someone pasted a link to resetting the DSRM pwd but I would assume that only applies going forward, not going backwards when I do not have access to it.