GoldenrozBook a call

Bare metal restore: backing up servers and legacy machines

A bare metal restore rebuilds a complete machine, including the operating system, applications, settings and data, onto hardware that has nothing installed. It requires an image-level backup of the system, not only copies of files. For servers and for old production-line PCs running DOS or Windows XP, it is the only practical way to recover from a dead disk, because reinstalling by hand is slow and often impossible.

A gold sphere under a clear glass dome on a navy base.

What a bare metal restore is

“Bare metal” means a computer with no usable operating system: a new server, a replacement disk, a machine wiped by ransomware. A bare metal restore takes that machine from empty to working in one operation.

The alternative is a manual rebuild. Install the operating system, find the drivers, install each application, locate license keys, reapply every setting, then restore the data. For a current server with documentation, that takes hours to days. For a fifteen-year-old PC that runs a cutting machine with software from a vendor that no longer exists, it may not be possible at all.

That is why an engineer asks one question about any backup: if this machine’s disk died now, what exactly would we restore from, onto what, and how long would it take.

Image backup vs file backup

File backup Image backup
What is copied Selected files and folders Whole disks or volumes, including boot data
Restores a deleted file Yes, quickly Yes, if the product can open images
Restores a dead machine No. You reinstall first Yes
Size Smaller Larger, reduced by compression and incremental images
Captures installed software and settings No Yes
Typical use User documents, shared folders Servers, workstations with special software, legacy machines

Most environments need both. File-level recovery handles the daily requests. Image-level recovery handles the bad day.

Windows Server includes this capability. Microsoft’s documentation for the wbadmin start backup command describes the -allCritical parameter as including all critical volumes, meaning those that contain the operating system’s state, and notes that it is useful when creating a backup for bare metal recovery. Third-party products offer the same idea with more scheduling, verification and central management.

Choosing disk imaging software

When you compare imaging products, check these points and ignore the feature count.

  • Boot media. It must be able to start the target machine and see its disk and network or USB controllers. For old hardware this is the deciding factor.
  • Offline imaging. The ability to image a disk while the installed system is not running.
  • Restore to different hardware or to a virtual machine.
  • Verification. The product should check that an image is readable after it is written.
  • Incremental images. So that daily backups do not copy the whole disk each time.
  • Open or documented format. You may need to read this image in ten years.
  • Old file system and partition support. FAT16, FAT32 and legacy partition tables for older machines.

The 3-2-1 backup rule

The rule is stated in a paper produced for US-CERT, “Data Backup Options”:

  • 3 copies of any important file: one primary and two backups
  • 2 different media types, to protect against different types of hazards
  • 1 copy stored off site

Applied to a server, that could be the live system, an image on a backup appliance in the building, and a copy in cloud storage or at a second site.

The same paper warns that rolling backups can silently carry corruption or malware from the primary files into the backup. That is the reason many organizations now add an offline or immutable copy that cannot be changed or deleted for a set period. If ransomware reaches the backup server with administrator rights, a copy it cannot alter is the one that saves you.

The rule covers where copies live. It says nothing about whether they work, which is the subject of the last section.

Backing up servers

For servers, bare metal capability is one layer of a design, not the whole of it.

  • Physical servers. Image the system volumes and back up data volumes on a schedule that matches how much work you can afford to lose.
  • Virtual machines. Back them up at the hypervisor level, which captures the whole machine and makes restoring to another host straightforward.
  • Databases and directory services. Use application-aware backups so that the data is consistent. A disk image taken while a database is mid-write can restore a damaged database.
  • Configuration. Keep exports of firewall, switch and hypervisor configuration alongside the images.

Set two targets per system: how fast it must be back (recovery time) and how much recent data may be lost (recovery point). Those targets decide the backup frequency and the restore method. Routine checks that backup jobs completed belong to server management services. The wider plan for losing a site, including order of recovery and who decides, is the scope of disaster recovery services.

The special case: production-line PCs on old operating systems

Factories, laboratories and warehouses often depend on a PC that nobody thinks of as IT. It runs a controller, a measuring instrument or a labeling line. It runs DOS, Windows 98, Windows 2000 or Windows XP because the machine it controls was built for that system. Microsoft’s lifecycle page shows extended support for Windows XP ended in April 2014.

These machines share a risk profile:

  • The disk is old and will fail
  • The software vendor may be gone, and installation media and license keys are missing
  • The software depends on specific hardware: an ISA or PCI card, a serial or parallel port, a license dongle
  • Replacement parts are bought second-hand
  • Nobody remembers how it was configured

They cannot be patched, so they should also be isolated on their own network segment with no internet access. The backup approach has four parts.

1. Image the disk offline

Do not install backup software on the old system. Schedule a short stop, shut the machine down and either start it from separate boot media that images the disk to external storage, or remove the disk and image it on another computer with a suitable adapter. Take a full sector-by-sector image so that unusual partitions and copy-protection data are captured. Make two copies and verify both.

2. Keep spare hardware

An image is useful only if there is something to restore it to. Old operating systems are bound to their disk controller and chipset, so the safest target is an identical machine. Buy one or two matching units, plus spare disks and any special interface cards. Restore the image to the spare and confirm that it boots and talks to the equipment. Label it and store it near the line.

A pre-imaged spare disk kept in a drawer turns a disk failure into a swap that takes minutes.

3. Virtualize where the hardware allows

Converting the old machine to a virtual machine removes the dependence on aging hardware and makes backup simple. It works when the software needs only standard devices. It becomes difficult when the software needs an internal card, precise timing or a hardware dongle. Serial and USB devices can sometimes be passed through to a virtual machine, and sometimes not.

Test the virtual copy against the real equipment during a planned stop before relying on it. Even when virtualization cannot replace the machine, a virtual copy is valuable as a reference for how the system was configured.

4. Document what the image cannot hold

  • BIOS settings, photographed screen by screen
  • Card positions, jumper settings and cable connections
  • Port assignments and device addresses
  • License keys and the location of any dongles
  • The steps to start the application and confirm it is working

This inventory, the images and the tested spare are the deliverables of our legacy system backup service.

Restore tests: the part most plans skip

A backup that has never been restored is an assumption. Test on a schedule and record the result.

  • Restore one file from each backup set every month
  • Restore a full server image to spare hardware or an isolated virtual machine at least twice a year
  • Boot the restored system and confirm that the application starts and the data is current
  • Time the restore and compare it with the recovery time target
  • Restore from the off-site copy, not only the local one
  • For each legacy PC, restore the image to the spare machine and run it against the real equipment during a planned stop
  • Confirm that the boot media still starts on the hardware you have
  • Check that restore credentials and encryption keys are stored somewhere that survives the loss of the systems being restored
  • Write down who performed the test, what failed and what was fixed

Encryption keys deserve a second mention. An encrypted backup whose key lived only on the server that died cannot be restored.

What to do next

Walk the production floor and the server room and list every machine whose loss would stop work. For each one, write down the date of the last image and the date of the last successful restore. Blank cells in the second column are the priority list.

We inventory, image and test legacy and server systems as a project through our legacy system backup and disaster recovery services. Projects start at $25,000 at a fixed price per phase. Ongoing backup monitoring and scheduled restore tests are part of managed IT services, a monthly retainer quoted after an assessment. For how off-site copies fit into a wider design, see what is hybrid cloud.

Questions buyers ask

What is a bare metal restore?

It is the recovery of an entire system onto a machine with an empty disk. The restore brings back the partitions, boot records, operating system, drivers, applications and data in one operation, so the machine starts as it was at the time of the backup.

What is the difference between an image backup and a file backup?

A file backup copies selected files and folders. An image backup copies whole disks or volumes, including the operating system and everything needed to boot. File backups are good for recovering a deleted document. Image backups are needed to recover a machine.

What is the 3-2-1 backup rule?

Keep 3 copies of any important data, on 2 different types of media, with 1 copy stored off site. The rule is described in a US-CERT paper on data backup options. Many organizations add an offline or immutable copy to protect against ransomware.

Can you restore a system image to different hardware?

Often, with care. Modern systems usually start on different hardware after drivers are added. Old operating systems are much less tolerant, because they are bound to specific disk controllers and chipsets. For legacy machines, plan on identical spare hardware or a virtual machine and test the restore before you need it.

How do you back up a Windows XP or DOS machine?

Shut the machine down, start it from separate boot media and take a full image of the disk to external storage, or remove the disk and image it on another computer. This avoids installing anything on the old system. Then restore the image to a spare disk or a virtual machine to prove it works.

Written by

CTO and technology partner

Shai leads the technology side: business software, SaaS and mobile apps, payments and escrow systems, Postgres security, AI agents built with Claude Code and MCP, and IT and infrastructure projects with senior infrastructure engineers.

Keep reading

Talk to the person who will do the work

One hour, no sales team. You talk to Daniel or Shai, we look at your account or your product, and you leave with a clear next step.

Or send the details

Book a 1-hour consultationWhatsApp us