Australian cloud for business software — built and run in Sydney since 2002

NetLeverage
NetLeverage

Windows Server Backup deleted older backups on a dedicated disk: how space is reused and what to check

Technical article
Backup & Recovery

About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.

When a disk is set up for scheduled Windows Server Backup, older versions are meant to disappear as the disk fills. Microsoft describes this as automatic disk usage management: when a new backup does not fit, the tool makes room by removing the oldest versions. If wbadmin get versions shows fewer dates than you expected, the usual explanation is that the disk, or the shadow copy storage on it, is smaller than the history you wanted. It is rarely corruption.

Prove that first. Then either give the backup more room, back up less, or send a second copy elsewhere. Do not delete shadow copies by hand to "tidy up".

What the symptom means

Microsoft's explanation, written for Windows Server 2008 and 2008 R2, is that each backup version is held in a volume shadow copy on the backup disk. After the data is written, the tool creates a shadow copy of the volume that stores the backup, and that copy is the "backup version". If the storage location is full, "Windows Server Backup automatically deletes the oldest backup version to make space for the current backup."

A second Microsoft post explains how. When a backup is running and the target lacks space, the tool shrinks the storage allocated to snapshots (the diff area). Shrinking removes one or more older snapshots, and so the versions that corresponded to them. It first checks that shrinking would free enough room. It will not shrink the diff area below one eighth of the target volume's size, "as we do not want to lose all past backups just to accommodate this one". That is why a backup can still fail for lack of space even with automatic management in place.

Microsoft's general Volume Shadow Copy Service FAQ says the same about any volume: when it runs out of space, "shadow copies for the volume are deleted, beginning with the oldest shadow copy."

Four limits on this behaviour matter:

  • The automatic rule applies to backups on a local disk. It does not apply to a network share. A share holds one version, which a failed run can overwrite, and that is a separate fault covered in "Windows Server Backup and wbadmin: the limits to know before you rely on them".
  • In Windows Server 2008, Microsoft says space for a system state backup was not managed automatically. System state versions have their own command, covered in "System state backups filling the backup disk: wbadmin delete systemstatebackup and -keepVersions".
  • Microsoft's 2012 overview lists "a deletion policy" for backups to a hard disk or volume: delete only when needed to make space, or delete after a certain number of backups. Microsoft's overview does not say where that setting is made, so do not assume a retention count is already set.
  • All of this is documented for 2008 to 2012 releases. The wbadmin pages for Windows Server 2016 to 2025 describe the commands but do not restate the automatic rule, so treat it as expected behaviour to confirm, not a guarantee.

Likely causes

Cause What you see How to confirm
The disk is smaller than the version history you want (normal space reuse) The oldest version moves forward after most runs; the newest runs complete Compare wbadmin get versions -backupTarget:<volume> dates over a week; check free space on the volume
Shadow copy storage on the backup volume has a low maximum Versions disappear well before the volume is full vssadmin list shadowstorage /for=<volume> and /on=<volume>: compare the storage reported with the volume size
Each run now changes much more data (database move, large import, re-encryption) History shortens from one week to the next with no setting changed Compare backup duration and shadow storage growth per run; ask what changed on the server
Shadow copies removed by something other than space reuse (a manual delete, a storage fault) Several versions vanish at once, on a night with no unusual change Compare shadow copy creation times from vssadmin list shadows with the version dates; read the System and Application logs around the time
An in-place upgrade or -allowDeleteOldBackups in the schedule Everything made before the upgrade is gone wbadmin enable backup with no parameters displays the currently scheduled backup settings; compare the upgrade date with the oldest version
The target is a network share Only one version ever exists The path in wbadmin get versions is a \\server\share location

Prove the cause before changing anything

These commands only read. Run them in an elevated prompt as a member of Administrators or Backup Operators.

  1. wbadmin get versions. With no parameters it lists the backups the local catalogue knows about, "even if those backups are not available". It can therefore show more versions than the disk holds.
  2. wbadmin get versions -backupTarget:<volume>, for example -backupTarget:H:. This lists what is on the disk. A disk that holds seven versions while the catalogue lists twenty points to space reuse.
  3. wbadmin get items -version:<version identifier> shows what one version contains.
  4. vssadmin list shadows /for=<volume> lists the shadow copies of the backup volume. The count and creation times should match the versions in step 2.
  5. vssadmin list shadowstorage /for=<volume> and vssadmin list shadowstorage /on=<volume> report where shadow copy storage sits and how much it uses.
  6. Get-WBSummary in PowerShell reports the next scheduled backup and the details of the last and last successful backups, which rules out failed runs as the reason for gaps.

A good result: the version count equals the shadow copy count, the newest version is last night's, and used space is stable. A bad result: a version count that keeps falling, shadow storage pinned at its limit, or last successful backup dates that stop.

Example (illustrative): fs01 backs up to a dedicated 2,000 GB disk. The first backup writes 1,200 GB. That leaves 800 GB for shadow copy storage. If each nightly run changes about 50 GB, each version costs roughly 50 GB of that storage, so about 800 ÷ 50 = 16 versions fit. After a data migration the nightly change rises to 100 GB, and the same disk holds about 800 ÷ 100 = 8 versions. Nobody changed a setting, yet history halved. The floor Microsoft describes, one eighth of 2,000 GB, is 250 GB. It limits how far shadow storage can shrink, not how many versions you keep. These figures simplify how a version uses space, and real use will differ.

Fix

More space or fewer items. The honest fix for a disk that is too small is a bigger one or a smaller job. Microsoft's enable backup page says more than one backup location can be added by running the command again with -addtarget. You can also reduce what is protected with -exclude.

Disruptive step: if you give a disk to wbadmin enable backup, Microsoft states "the disk will be formatted before use, and any existing data on it is permanently erased". There is no way back. Check the disk identifier with wbadmin get disks and make sure it holds nothing else.

Shadow copy storage limit. If vssadmin list shadowstorage shows a low limit, vssadmin resize shadowstorage /for=<ForVolumeSpec> /on=<OnVolumeSpec> /maxsize=<MaxSizeSpec> raises it. Microsoft's page warns: "Resizing the storage association may cause shadow copies to disappear." If no /maxsize is given, there is no limit. Microsoft's 2011 post names the graphical route as the volume's properties, Shadow Copies, Settings.

Risk: resizing can remove versions you still need. Take a copy of the backup volume's contents to another disk first, or choose a time when losing the oldest version is acceptable.

Keeping a set number of versions. Remove-WBBackupSet removes backups from a target. Its -KeepVersions parameter "specifies the number of the most recent backups to keep", and -DeleteOldest removes the oldest. Microsoft's page does not say whether this frees disk space immediately, so check with vssadmin afterwards.

Disruptive step: Remove-WBBackupSet -Force deletes without asking. A deleted version cannot be recovered. Use -KeepVersions only after you have confirmed the number you pass is more than you need.

Do not delete shadow copies by hand. Microsoft's 2011 post says that deleting backups other than system state ones by removing their shadow copies is "not recommended", that incorrect deletion is permanent, and that "deleting backup versions in middle would not guarantee that free space gets created".

Upgrade history. If -allowDeleteOldBackups is in the schedule, understand what it does before the next upgrade. Microsoft's description is that it "overwrites any backups made before the computer was upgraded".

Second copy. Space reuse is by design, so a longer history needs a second target, not a bigger hope. Add a second disk target, or copy completed backups elsewhere, such as an offline disk.

Confirm the fix

  • After several runs, wbadmin get versions -backupTarget:<volume> should list the number of versions you planned for, and the oldest date should stop moving forward.
  • The shadow copy count from vssadmin list shadows /for=<volume> should equal the version count.
  • Get-WBSummary should show a last successful backup that matches the schedule.
  • A listed version is not a tested restore. Restore a sample file or folder from the oldest version you keep.

What to watch afterwards

  • Record the oldest version's date each week. When it stops being at least your target retention old, act before it matters.
  • Plan disk size as the first full backup plus retention in nights multiplied by the typical nightly change, plus headroom. For example, 1,200 GB + (30 × 50 GB) = 2,700 GB before headroom, on figures you replace with your own.
  • After any bulk data change on the server, expect the history to shorten.
  • Watch for failed runs as well. A run that fails leaves a gap in history that no amount of disk space repairs.

Risks and limits

  • These commands have not been run or lab-tested.
  • The space reuse rule and the one-eighth floor are described by Microsoft for Windows Server 2008 and 2008 R2. Microsoft does not restate them for current releases.
  • Microsoft's 2012 overview mentions a deletion policy but does not show where it is set.
  • A user on Microsoft Q&A reported all versions being deleted after a cumulative update on Hyper-V Server 2019. The only reply quotes the 2011 post. This is a community report, and no cause is documented.
  • The one-eighth floor is a lower bound on shadow storage, not a promise that one eighth of the disk will always hold history.
  • Manual removal of versions is permanent, and the disk formatting step at setup erases existing data.

Sources