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.
A defragmentation run moves the clusters of files from one place on the disk to another. A block-level backup lists what changed by location on the disk, not by what the file contains, so moved blocks are backed up again even though no data changed. Shadow copies work at the same level: each block that is rewritten has to be preserved first. Arcserve calls the backup-size effect expected behaviour. Loss of shadow copies after a defragmentation is documented only for Windows Server 2003 and older Windows versions. Both effects are largest when Windows runs its weekly Optimize Drives task during or just before the backup window.
The fix is to know when the task runs, keep it away from the backup (and, ideally, from heavy shadow copy use), and give shadow copy storage room to absorb a defrag run.
What the symptom means
Microsoft's developer documentation describes defragmentation as "the process of moving file clusters on a disk to make them contiguous". Nothing in a file's contents has to change for that to happen.
Two documented mechanisms turn that into growth:
Block-level backups. Arcserve's documentation says that a volume defragmentation by the Windows native tool "affects the size of the block-level backups because Arcserve UDP continues to incrementally back up all changed blocks. It means that blocks that shifted during the defragmentation are included in the backup, even if no data has changed in the files... This behavior is expected." Acronis says that "an incremental or differential backup created after the disk is defragmented will be considerably larger than usual", because the defragmentation program "changes file locations on the disk, and backups reflect these changes." Both pages describe block-based products. A backup that reads files by name and modified time, and not blocks, is described by neither page; that it would be unaffected is reasoning, not a statement from either vendor.
Shadow copies. Microsoft's Volume Shadow Copy Service page explains that the system provider uses copy-on-write: "each block to be modified is read and then written to the volume's shadow copy storage area (also called its diff area)". It adds that "if there are many changes, the copy-on-write method can become expensive." A defragmentation run produces a great many block writes at once. That a defrag run therefore needs diff area space in proportion to the blocks it moves is reasoning from those two statements, not a Microsoft figure.
Microsoft's developer page on defragmenting files asks defragmentation software to align moves in 16-kilobyte steps: "This reduces copy-on-write overhead when shadow copies are enabled", and "a volume formatted to a 16-KB cluster size will result in all move files being optimized". It does not say how the built-in defrag tool moves data, so do not assume the built-in tool avoids the overhead.
A Microsoft support article states that "a similar problem is caused by defragmenting a volume with small cluster size, where all shadow copies are deleted", and that the VolSnap event ID 25 is logged. That support page is a hotfix note for Windows Server 2003, so treat it as the origin of the behaviour and not as a statement about current Windows. A third-party defragmenter's manual (O&O Defrag) reports the same for Windows Vista, 7, 8, Server 2008 and Server 2012: a complete defragmentation of drives that are imaged by shadow copies "may lead to losses of old image sets", and says that on Windows XP64 and Server 2003 the problem does not occur with a cluster size of 16 KB or larger. That is a vendor's statement about its own product's interaction, not a Microsoft table for current versions.
When does Windows defragment on its own?
Microsoft's defrag page says the defragmentation process "runs scheduled task as a maintenance task, which typically runs every week", and that an administrator can change how often it runs in the Optimize Drives app. For SSDs the scheduled run does traditional defragmentation and retrim "once per month". Defrag also needs free space: "A volume must have at least 15% free space for defrag to completely and adequately defragment it". Below that it only partly defragments.
What Optimize-Volume does with no parameter depends on the drive type. For HDD, fixed VHD and Storage Space it runs -Analyze -Defrag. For an SSD with TRIM support it runs -Retrim. For thinly provisioned storage, dynamic VHD and differencing VHD it runs -Analyze -SlabConsolidate -Retrim. Microsoft's pages do not say which of these operations the weekly scheduled task runs on a virtual disk, so check on your own server what it did and when.
Likely causes
| Cause | What you see | How to confirm |
|---|---|---|
| Weekly Optimize Drives task ran a traditional defrag on a disk type that gets one | Large incremental on the day after the task, then normal sizes | Optimize Drives "last run" time; task run-time information; backup sizes by date |
Manual defrag or Optimize-Volume -Defrag, including free space consolidation (defrag /x) |
One-off large incremental after a maintenance change | Admin session history, change record |
| Shadow copy storage too small for the blocks moved | Shadow copies missing after the run; VolSnap event logged | vssadmin list shadows shows fewer copies than before; event log (see the related article on event 25) |
| Cluster size below 16 KB on a volume with shadow copies | Shadow copies lost after defragmentation (documented for Server 2003; O&O reports older versions) | Volume allocation size compared with 16 KB |
| A restore onto the volume before the backup | Next incremental larger than usual | Restore log and date; Acronis states a restore changes file locations on the disk |
| Backup change tracking was reset at the same time | Next backup is a full read, not just moved blocks | Backup job log; compare with the related article on change tracking resets |
Prove the cause before changing anything
These checks read or analyse only. Nothing on the volume moves.
-
Match dates. Take the date and time of the large incremental from your backup product's job history. Then find when Optimize Drives last ran. The defrag page refers to the "Last run time" shown in the Optimize Drives app; open the app and read it for each volume.
-
Find the task and its run time. The task is a scheduled task; the name is not given on Microsoft's pages, so filter by a word. This filter is a choice made here, not a Microsoft-documented name:
Get-ScheduledTask | Where-Object TaskName -like '*defrag*' | Get-ScheduledTaskInfo
Microsoft documents Get-ScheduledTask, Where-Object and piping task objects to Get-ScheduledTaskInfo, which "gets the last run-time information for a scheduled task". Compare the last run time with the backup's size jump.
-
See what Optimize would do to each volume.
Optimize-Volume -DriveLetter D -Analyze -Verbose"reports only the current optimization state". Run it per volume.defrag D: /a /vdoes the same through the command line; Microsoft describes/aas analysis and/vas verbose output with fragmentation statistics. -
Check shadow copy use.
vssadmin list shadows /for=D:
vssadmin list shadowstorage /for=D:
The first lists the shadow copies of the volume. The second "reports the disk space that storage uses". If copies that existed last week are gone and a defrag ran in between, that fits. For events, use the related article on VolSnap events.
- Read the volume's allocation size. Microsoft's Win32_Volume class documents a
BlockSizeproperty ("Size in bytes of the blocks in this storage extent"):
Get-CimInstance -ClassName Win32_Volume | Select-Object DriveLetter, BlockSize
Microsoft's page does not use the word cluster for this property; read the figure with that caution. 16 KB is 16,384 bytes.
- Rule out a reset. If the backup log says the product reread the whole disk or reset change tracking, the cause is not defragmentation alone. See the related article on change tracking resets.
Example (illustrative): fs01 has a 1,000 GB data volume D: with 800 GB used. Nightly incrementals run about 8 GB (1% of 800 GB). The Sunday maintenance task runs a defragmentation that moves about 25% of the data, 200 GB. Monday's incremental is about 208 GB, 26 times normal (208 / 8 = 26). With 28 nights kept, four weekly defrag runs sit in the retention window: 28 × 8 GB = 224 GB of ordinary changes, plus 4 × 200 GB = 800 GB of moved blocks, a total of 1,024 GB, or about 4.6 times the 224 GB that ordinary changes would need (1,024 / 224 = 4.57). All figures are invented to show the arithmetic, and a backup product that stores only data new to its store may keep less.
Fix
Cause: the weekly task runs inside the backup window
Change when and how often the task runs, in the Optimize Drives app, as the defrag page describes. Put it well away from the backup, and where possible directly before a full backup, so that the moved blocks fall into data you were going to capture anyway. Acronis suggests rescheduling or disabling defragmentation, or taking a new full backup after defragmenting and before more incrementals.
Risk: switching the Optimize Drives task off entirely also stops retrim (on SSD or thin disks) and slab consolidation, which the Optimize-Volume documentation lists as the default operations for those drive types. Prefer a new schedule over disabling it. Check with your storage or virtualisation provider whether a traditional defrag brings any benefit on your disks, because it may not on every platform.
Risk: moving the schedule changes when defragmentation and retrim run, and a new time can collide with other night work or land inside a different backup's window. Note the current schedule in the Optimize Drives app before changing it; to back out, set the noted schedule again.
Cause: shadow copy storage too small
Enlarge the maximum shadow copy storage for the volume, or move it to another volume, using the steps in the related article on VolSnap events. The command is:
vssadmin resize shadowstorage /for=D: /on=D: /maxsize=<size>
Risk: Microsoft's own warning on this command is "Resizing the storage association may cause shadow copies to disappear." Do it outside a backup and when you can accept losing the existing shadow copies.
Cause: cluster size below 16 KB with shadow copies
A volume formatted with a 16 KB cluster size removes the alignment problem Microsoft's developer page describes. That requires formatting.
Risk: reformatting a volume erases its data. There is no way back except a restore from a backup. Treat this as a design choice for a new volume, not a fix for a volume that is live.
After the event
Let the next incrementals run and confirm sizes fall back (see below). If you take a new full backup, do so on a normal schedule, since Acronis advises one after a defragmentation.
Confirm the fix
- The next scheduled run of Optimize Drives falls outside the backup window; the app shows the new schedule.
- The incremental after the next run is within the normal range for the volume, and not in the tens of times higher, or the cause is elsewhere.
vssadmin list shadows /for=D:shows the same number of copies after the run as before.- The volume has at least 15% free space, so that the defrag finishes properly when it runs.
What to watch afterwards
- Backup size by date, with the Optimize Drives run dates beside it.
- Shadow copy count and storage use after each maintenance window.
- Disk type changes (physical to virtual, HDD to SSD, thin provisioning), because the default Optimize operation changes with them.
- Restores onto the volume, which Acronis says change file locations and make the next incremental larger.
Read next
Related articles you may want next.
- Backup & Recovery'The shadow copies of volume were deleted': volsnap events 25, 33 and 36 and shadow copy storage that cannot growVolsnap event 25, 33, 35 or 36 and shadow copies deleted: what each event means, vssadmin checks, and how to resize or move shadow copy storage safely.Read article
- Backup & RecoveryPrevious Versions and shadow copies are not a backup: what they are good for, and where they stopUnderstand local Windows shadow-copy dependencies, automatic deletion, administrative access and why independent retained backups still matter.Read article
- TroubleshootingIncremental backups suddenly run as full after a CBT resetInvestigate full reads after CBT changes, distinguish configuration from integrity, and understand supplier-supported recovery and restore acceptance.Read article
Risks and limits
- These commands have not been run or lab-tested. They are taken from Microsoft Learn pages and vendor documentation.
- Microsoft's shadow copy and cluster size statements that name defragmentation come from a Windows Server 2003 support article and from the developer documentation. No current Microsoft page says what the built-in defrag tool does to shadow copies on Windows Server 2016 to 2025.
- The O&O Defrag statements describe a third-party tool and older Windows versions.
- Arcserve and Acronis describe their own products; other products track changes in different ways and with different block sizes, so the size of the effect varies.
- The name and settings of the scheduled task are not given on Microsoft's pages; filter by name on your own server.
- The size of a defrag effect depends on how much data the run moves, which no source gives; the example's 25% is invented.
Sources
- defrag — Microsoft Learn, last updated 1 November 2024, accessed 17 November 2026
- Optimize-Volume — Microsoft Learn, page metadata 14 May 2025, accessed 17 November 2026
- Defragmenting Files — Microsoft Learn (Win32 apps), last updated 10 September 2024, accessed 17 November 2026
- Volume Shadow Copy Service (VSS) — Microsoft Learn, last updated 7 July 2025, accessed 17 November 2026
- Time-out errors occur in Volume Shadow Copy service writers, and shadow copies are lost during backup and during times when there are high levels of input/output — Microsoft Support, Windows Server 2003 article, accessed 17 November 2026
- vssadmin list shadows — Microsoft Learn, accessed 17 November 2026
- vssadmin list shadowstorage — Microsoft Learn, last updated 8 September 2026, accessed 17 November 2026
- vssadmin resize shadowstorage — Microsoft Learn, last updated 1 November 2024, accessed 17 November 2026
- Win32_Volume class — Microsoft Learn (legacy documentation), accessed 17 November 2026
- Get-CimInstance — Microsoft Learn, page metadata 16 January 2026, accessed 17 November 2026
- Select-Object — Microsoft Learn, page metadata 11 August 2026, accessed 17 November 2026
- Get-ScheduledTaskInfo — Microsoft Learn, accessed 17 November 2026
- Review the Prerequisites and Considerations (Arcserve UDP 7.0 solutions guide) — Arcserve documentation, no date shown, accessed 17 November 2026
- Incorrect size of incremental or differential backup archives — Acronis knowledge base, last updated 11 December 2025, accessed 17 November 2026
- Conflicts with shadow copies during defragmentation — O&O Software, O&O Defrag 18 manual, no date shown, accessed 17 November 2026
- Get-ScheduledTask — Microsoft Learn, accessed 17 November 2026
- Where-Object — Microsoft Learn, accessed 17 November 2026
This article was originally published on .