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 synthetic full is not read from the virtual machine. The backup software builds a new full file on the repository from the previous full and the increments that follow it. So when it fails, the causes Veeam's documentation points to are on the repository side: not enough room for another full-sized file, block cloning that is not active, a retention rule that keeps an old chain longer than you expected, or a schedule rule that cancels the job.
Do not delete backup files by hand to "make room". In a forward incremental chain every increment depends on the files before it, and removing the wrong one can leave restore points that cannot be used.
What the error means
The description below follows Veeam Backup & Replication 13, because its user guide sets out the synthetic full process in detail. Other products use the same idea with different rules, so read your own product's guide for the exact behaviour.
According to Veeam's guide, a synthetic full backup session works in this order:
- On the scheduled day the job starts a session and first takes an ordinary incremental backup from the production storage. That keeps the new full up to date with the machine.
- At the end of the session, the Veeam Data Mover on the repository builds a new full file from files already in the chain, including that new increment.
- When the new full exists, the Data Mover deletes the increment created at the start of the session.
- The new full resets the chain. Later increments use it as their starting point. The previous full "remains on disk until it is automatically deleted according to the retention policy".
The guide says the result "contains the same data you have if you create an active full backup". Nothing is read from the source for the full itself, which is why the method is gentle on the network and the production host, and why the repository takes the whole load.
The guide does not describe what a half-built full looks like after a failure. Treat any unexplained large file in the job folder as unknown until you have matched it against the restore points your console lists.
Likely causes
| Cause | What you see | How to confirm |
|---|---|---|
| The repository cannot hold another full-sized file | The synthetic session fails; read the job log for the reason. Free space is below the size of the latest full | Compare free space with the size of the newest full file (steps below) |
| Fast Clone is not active, so the new full is a real copy | The synthetic full is slow and the repository grows by close to a full file each time | Check file system and repository type against Veeam's Fast Clone conditions |
| A retention rule keeps the old chain | Space looks wasted. More restore points exist than the retention setting says | Compare the restore points listed with the rule in Veeam's forward incremental retention page |
| Schedule rules cancel the synthetic full | No synthetic full appears and nothing failed | Look for an active full on the same day, or a chained job whose first job did not run |
| The target is object storage without a long-term policy | No synthetic full appears; Veeam says synthetic fulls "cannot be created independently" for object storage | Check whether the GFS policy is enabled for that job |
| An SMB or Windows repository lacks a Fast Clone requirement | Fast Clone is not used although ReFS is in place | Check SMB 3.1.1, ReFS 3.1 and the two required control codes in the guide |
Prove the cause before changing anything
All of these checks only read information.
1. Free space against the size of a full. On a Linux repository, Veeam's own configuration steps use this command to check a disk:
df -hT
It lists each file system, its type and the space used and available. Note the available space for the volume that holds the job folder. Then read the size of the newest full file (the .vbk file) in the job's folder, or in the console under the backup's disk view. If available space is smaller than that file, a real-copy synthetic full cannot complete. Veeam's guide for compacting a full file says the repository "must have enough space to store a file of the full backup size", and a synthetic full builds a file of that size too. Veeam's guide does not state a space rule for the synthetic full itself, so this is a reasoned comparison and not a documented threshold.
2. File system and Fast Clone conditions. For Linux, Veeam lists: an XFS file system, a supported distribution, a kernel with reflink support, and a block size of 1 KB to 4 KB with checksums on. For Windows and SMB repositories it lists Windows Server 2016 or later (or Windows 10 Pro for Workstations or later) with ReFS 3.1 or later, and for SMB shares the SMB 3.1.1 protocol plus support for FSCTL_DUPLICATE_EXTENTS_TO_FILE and FSCTL_SET_INTEGRITY_INFORMATION. If any item is missing, expect real copies.
3. History of the chain. Veeam says that after you move a chain onto a Fast Clone repository, or enable Fast Clone on an older repository, the chain does not benefit until an active full backup has run. Veeam's Fast Clone page also names a compact operation as an alternative, but Veeam's compact page says compact "works for forever forward incremental backup chains or reverse incremental backup chains" and that for it "you must not schedule active or synthetic full backups". A job that builds synthetic fulls therefore uses the active full, not compact. If the chain was moved, that is a documented reason block cloning is not in use. Veeam adds that an active full is not required after a move if you use the Veeam copy or move features.
4. Extents. If the repository is an extent in a scale-out repository, Veeam says Fast Clone "will not be used" when backup files sit on different extents, and recommends the Data Locality placement policy.
5. Schedule. Read the job's schedule. Veeam documents that a synthetic full is not created if an active full is scheduled for the same day, and that a chained job gets no full on days when its first job does not run.
Fix
Not enough space. Add capacity, shorten the retention, or move the synthetic day to a time when the oldest chain is about to expire.
Risk: shortening the retention makes Veeam remove older restore points after the change, and they cannot be recovered afterwards. Impact: you can no longer restore to those dates. There is no way back, so check which points you still need, and keep a copy of anything required, before you change it. Adding capacity or moving the day removes nothing.
Do not delete chain files by hand. Veeam's retention guide states that "if you delete a full backup file, the entire chain of incremental backup files will become useless", and deleting an increment makes every later restore point unusable.
Fast Clone not active. Correct the file system or repository setting first. Then run an active full backup so the chain starts using block cloning, as Veeam describes. Do not schedule the compact operation on a job that builds synthetic fulls; Veeam's compact page says compact is for forever forward and reverse incremental chains and that you must not schedule active or synthetic fulls with it.
Before you start the active full, check free space against the size of one full file. The active full writes another full-sized file to the repository, which is the same shortage that makes a synthetic full fail; in the example below only 2.05 TB is free against a 3.0 TB full. Free the space first, or the active full fails for the same reason.
Disruptive step: an active full reads the whole machine from production storage and sends it across the network. It adds load to the host and the network during the run. It adds a new full rather than altering the old chain, so the existing restore points stay.
Chain moved by hand. Same step as above: an active full.
Schedule conflict. Move the active full or the synthetic full to a different day, or stop the chained job from skipping the day. Run the changed schedule and watch the next scheduled day.
Object storage. Veeam says synthetic fulls "cannot be created independently" for jobs that target object storage and that, to include them, "you must enable the GFS policy".
Changing the backup method. Veeam says switching methods does not convert the old chain: it "creates a new backup chain next to the existing one". That temporarily needs space for both chains.
Example (illustrative): repository backup01 has a 10 TB volume. One machine, fs01, has a 3.0 TB full file and about 0.15 TB of new data a day. Synthetic fulls run weekly, and retention is seven restore points. The oldest chain is still kept (as in Veeam's retention example, where an old chain waits for its last increment to become outdated). At the start of the synthetic day:
- Old chain: 3.0 TB full + 6 increments of 0.15 TB = 3.0 + 0.9 = 3.9 TB.
- Current chain: another 3.0 TB full + 6 increments = 3.9 TB.
- Total used: 3.9 + 3.9 = 7.8 TB. Free: 10 - 7.8 = 2.2 TB.
- The job first adds a 0.15 TB increment, so 2.05 TB is free.
- Without block cloning the new full needs about 3.0 TB. 2.05 TB is less than 3.0 TB, so the session fails.
With block cloning active, Veeam says data blocks are referenced and only copied when files are modified, so the new full would need far less. The same job can fail on one repository and pass on another, depending only on the file system.
Confirm the fix
- The next scheduled synthetic day produces a new full restore point in the console, and the temporary increment from the start of the session is gone.
- Free space on the volume after the session stays above the size of one full file.
- Read the job session log for the synthetic phase and check it finished without errors or warnings.
- If you changed the schedule, check that the synthetic full appears only once that day. Veeam creates the file once a day and runs ordinary increments for any further runs.
- Run the product's health check. Veeam describes it as a CRC check for metadata and a hash check for data blocks of the latest restore point, with a default schedule of 5:00 on the last Saturday of the month. A restore test of a file or a machine is the stronger check, because the health check covers only the latest restore point.
What to watch afterwards
- Free space on the repository against the size of one full file, tracked weekly. A graph of free space that dips toward zero on synthetic days warns you before a failure.
- The number of restore points on disk against your retention setting. In forward incremental chains there are more points than the setting says on some days, because the old chain is kept until its last increment expires.
- The effect of any change to the repository: a new volume, a moved chain, a changed extent policy. Each can switch block cloning off.
- Whether the synthetic day still lines up with other jobs. A new chained job or a new active full schedule can silently cancel the synthetic full.
Read next
Related articles you may want next.
- Backup & RecoverySize Backup Retention Using Measured Storage GrowthBuild a capacity forecast from stored growth, copy formats and retention dependencies, with explicit headroom, recovery needs and uncertainty.Read article
- Backup & RecoveryReading a backup report: what success, warning, partial, failed and missed each mean for recoveryWhat backup report statuses mean for recovery: success, warning, partial, failed and missed, using AWS, Azure and Veeam documentation, with a triage rule.Read article
- Backup & RecoveryFinding a bad backup before the day you need itWhat hashes, signatures and sample restores detect in backups, which faults they miss, and when to treat a copy as unusable.Read article
Risks and limits
- The Veeam user guide describes the behaviour of build 13.1.2.57. Older builds, other Veeam products and other vendors differ.
- Veeam's guide does not describe the state left behind by a failed synthetic session, and it does not give a numeric space threshold for building a synthetic full. The arithmetic above is a reasoned estimate, not a vendor figure.
- Deleting backup files outside the product can destroy a chain. Use the product's own delete and retention actions.
- Block cloning on ReFS has limits of its own: Microsoft lists, among others, that source and destination files must be on the same ReFS volume, and that the number of file regions that can map to the same physical region has a maximum of 8,175.
- A health check and a successful job do not prove a restore will work. Only a restore does.
- These commands have not been run or lab-tested.
Sources
- Synthetic Full Backup, Veeam Backup & Replication User Guide for VMware vSphere — Veeam, page updated 2025-07-09, accessed 17 November 2026
- How Synthetic Full Backup Works, Veeam Backup & Replication User Guide — Veeam, page updated 2025-03-20, accessed 17 November 2026
- Fast Clone, Veeam Backup & Replication User Guide — Veeam, page updated 2026-07-20, accessed 17 November 2026
- Forward Incremental Backup Retention Policy, Veeam Backup & Replication User Guide — Veeam, page updated 2026-01-12, accessed 17 November 2026
- Compact of Full Backup File, Veeam Backup & Replication User Guide — Veeam, page updated 2025-08-13, accessed 17 November 2026
- Switching Between Backup Methods, Veeam Backup & Replication User Guide — Veeam, page updated 2025-08-12, accessed 17 November 2026
- Health Check for Backup Files, Veeam Backup & Replication User Guide — Veeam, page updated 2026-08-04, accessed 17 November 2026
- Block cloning on ReFS, Microsoft Learn — Microsoft Learn, last updated 2024-11-01, accessed 17 November 2026
This article was originally published on .