Backup strategies - MS Server Primary Domain Controller (or SDC)

Good afternoon.

I have a virtual environment running on ESXi and HyperV. Around 90 VMs over 5 hosts.

We use windows server 2022 (a few other machines older). Thin clients use LTSC W10 (still have some older clients and we use some touch screen computers as thin clients as well)

Some of the systems have Symantec backup (now trelix) backups configured.

the ESXi hosts are all backed up by a Synology NAS using Active Backup (very neat solution) with 10G networking this is quite handy. It only looks for changes so it runs daily.

The NAS itself has a backup to another building (second Synology NAS that is not part of the domain and networked independently) this uses hyperbackup.

Our final tier of backup is to shut down each VM and export it onto a hard disk which is then sent off site.

my question is around the domain controllers -are these two methods (Synology and exporting the VM in a shut down state) acceptable or good practice?

If i ever have to restore the domain controller - what potential issues will i face?

Other comments or questions welcome - :slight_smile:

You should be fine to export a shut-down VM, it’ll just resync when reconnecting. Worst case you demote the DC and rebuild it.

Assuming that snapshots are atomic for the backups, the same is the case for the Synology backup. The DC should boot, run chkdsk to clean the disks, then restart and integrity check its AD database and sync back up to the rest of the domain. Assuming no issues during chkdsk, you should be fine.

You may want to test this though - I’m assuming that there are atomic snapshots in play here. (And also assuming that there’s no long delay between the backup and it being used.)

I do have something to note about your environment in general: please think about how you start up from cold, and whether or not your domain controllers are accessible.

Scenario: A site goes dark due to power issues. When restarting, how does it all work? Specifically, how do your virtualisation hosts find things?

AD is probably your DNS. If it’s only a guest on the network, then how does a machine do DNS queries before the guest can boot?

I’ve watched in horror as everyone realised that there was no DNS because it was only there as guests - and the hosts couldn’t do DNS queries to find their storage or to join together as clusters, so they just stopped as a failsafe. No storage, no clusters, no guests, nothing of any actual use. Manual editing of network configurations was needed to get it all working.

Fortunately you have two sites. The best, cheapest way is to double-check all of your network devices, storage devices and hosts network settings and make sure that their DNS servers straddle both sites. The likelihood of having to cold start AND having no WAN link between the sites is low.

But it’s not zero. A tiny 1U pizza-box “emergency” AD server at each site would be quite justifiable if you really want to cover your cold-start options. It can be something you only point servers at, so it doesn’t need to be big.

It’s great that you’re looking at your backups. The next step is to test them.

The step after that is to start planning for that cold start of a datacentre that you never, ever want to actually have to do. :wink: