Ways to quickly restore a server system on new hardware?

Question

System administration

Please, give some practical articles or tell us about your personal experience of quickly restoring server (MS) operating systems (files, printers, AD) on new hardware, incl. virtual: solutions, pros, cons. Ideally, you need a solution for a remote office for 15-17 people without a permanent admin, and which at least once can independently recover and earn on backup hardware (or with the help of a minimal participation of a non-specialist on site).
There is an understanding of how to do this, but there is no understanding of the pros and cons of solutions, prices and hardware requirements. I'm afraid that without prior information I can spend a lot of money for nothing.

Answer 1

In terms of cost, I would suggest focusing on fast recovery rather than backing everything up.
On some usually system unit, a tftp server and a prepared image are configured, with the ability to remotely connect to it and the necessary tools.
A configuration management system is used to set up the entire architecture (Puppet, Chief, Ansible, PowerShell DSC to taste), and the actual scripts are stored on the same PC.
If you encounter problems that are not related to the death of iron, you will need to turn on this configured PC and restart the Server. It is loaded from a tftp server and you connect to it - fix everything or reboot the system and feed configuration scripts to restore the infrastructure. Check that everything is working.
In this situation, it is necessary to ensure the security of the data so that the reinstallation of the system does not affect them in any way, but it is not necessary to buy 2-3 servers as in the redundancy and High availability schemes. But you have to study a lot in order to organize the system itself.

Answer 2

In the case of server virtualization, there are no recovery problems if you build the right virtual infrastructure: shared external storage for partitions with VM disks, 2 identical servers for hypervisors (servers can boot from a flash drive or memory card), both are connected to storage. In storage, mirror disks + 1/2 hot-swap disks.
In this case, recovery comes down to simply starting the VM on another physical server.
If you raise HA, then recovery will occur automatically.
If a disk in the storage fails - the data will not be lost thanks to the mirror, the failed disk is automatically replaced by a hot-swappable disk. After the incident is detected, all that remains is to replace the failed disk with a new one and define it as a hot swap.
But, usually, such a solution to support a network of 20 machines is too expensive, but it is really reliable and quite easy to maintain.
The bottleneck is the storage, but even here there are options, though with an increase in the cost of the solution.

Answer 3

Three hypervisors in a cluster, albeit with local storage (Storage Spaces Direct - but expensive as a dog), or with shared iSCSI storage. As an option, consider Nutanix Community Edition because it is free, but it asks for hardware one and a half times more than the total VM requirement, and three times for storage, plus it asks for mandatory SSDs for metadata. There are also three hosts (maximum 4), but unlike S2D, they also live on gigabit, albeit slowly. By the way, they are administered by an engineer of any training who knows how to steer any hypervisor, but there are a lot of pitfalls, mainly regarding mandatory updates and the use of NVMe.