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.
"Forever incremental" means the source is read in full once. After that, only changes are sent. The word hides a second question: what does the backup store do with all those changes so the chain does not grow without end? Three well-documented products answer it three different ways, and "always full" is a fourth label that describes what you see at restore time rather than what travels each night.
Knowing which mechanism sits behind the label tells you how much work the backup store does each day, what a restore reads, and what a damaged piece can affect.
What does "forever incremental" mean?
After the first full backup, every run sends only data changed since the run before. Taken alone that would produce an ever-longer chain of increments, and a restore would need every one of them. So every product that uses the label also has a rule that keeps the chain short. The rule is where the products differ.
How do the three documented mechanisms differ?
| Label and product | What is sent each night | What the store does afterwards | What a restore reads |
|---|---|---|---|
| Forever forward incremental (Veeam) | Changed blocks, saved as an increment file | When a point expires, the oldest increment's data is injected into the full file, and the increment is removed | The full plus the increments up to the chosen point |
| Always Incremental (Bareos) | Changed files only | A separate consolidation job merges old increments, and by default the full, into a new full on the server | The consolidated full plus the newer increments |
| Incremental upload, every snapshot a full (Proxmox Backup Server) | Only chunks the server does not already hold | Nothing to merge: each snapshot lists every chunk it needs | One snapshot, which references all its chunks |
Veeam describes its chain as "the first full backup file (VBK) and a set of forward incremental backup files (VIB) following it". To meet the retention setting it "injects data of an incremental file into the full backup file before deleting the increment", so "the full backup file 'moves' one step forward in the backup chain" (Backup Methods; Forever Forward Incremental Backup Retention Policy).
Bareos describes it as "Client Backups are always run as incremental backups", with consolidation of increments "older than a configurable age ... into a new backup". It adds a limit: "Always Incremental backups are only suitable for file-based backups" (Always Incremental Backup Scheme).
Proxmox Backup Server says "the actual upload of Snapshots is incremental while each Snapshot references all chunks and is thus a full backup" (Technical Overview). Its FAQ repeats that "each backup still references all data and such is a full backup" (FAQ).
What is "always full", then?
It is a statement about restore, not about transfer. When a provider says every backup is a full, one of two things is true:
- Each backup is a complete copy that was assembled from earlier data. Veeam's synthetic full is an example: it "synthesizes a full backup from data you already have in the backup repository".
- Each backup is a list of pieces, and unchanged pieces are shared with earlier backups, as in the Proxmox design above.
In both, the night's transfer is still small. The label tells you a restore needs no chain of increments, not that a whole copy was resent or stored again.
What does the merge cost?
Merging is not free. It moves the cost from the network to the backup store.
Veeam notes that injecting increments into the full "can lead to the fragmentation of the full backup file", so a periodic compact is scheduled. The compact "creates a new full backup file in the target repository", and the repository "must have enough space to store a file of the full backup size" while it runs (Compact of Full Backup File). On a deduplicating storage appliance, the retention page says Veeam "appends data from the first incremental backup file in the chain to the full backup file" instead of reusing empty blocks, so the chain "may consume more disk space".
Bareos is blunt about its basic scheme. Each day "the full backup is consolidated with the oldest incremental backup, which means that more or less the full amount of data being stored on the client is moved". The consolidation runs "locally on the storage daemon without client interaction", yet it "is still an enormous amount of data being touched". Its setting Always Incremental Max Full Age leaves the full untouched until it reaches that age, and Max Full Consolidations limits how many consolidations that include the full a job starts.
Example (illustrative): fs01 holds 1,000 GB and changes 30 GB a day. Each night only 30 GB leaves the server. Under a Bareos-style basic always-incremental scheme the consolidation then handles roughly the 1,000 GB full again on backup01, so the network moves 30 GB a night while the store works through about 1,000 GB. Under a Veeam-style forever forward chain, the store merges about 30 GB into the full each night, plus the periodic compact. With the chain kept at a full plus six 30 GB increments, the steady stored size is 1,000 + 6 x 30 = 1,180 GB. The figures show the shape of each design, not measured results.
Which restore point is fastest?
With a Veeam forever forward chain the oldest point is always a full, so it is the fastest: "Restore to the earliest restore point ... is the fastest compared to other methods because the first available restore point is always a full backup" (Backup Methods). Veeam says restores to the other points "can be compared by speed with the FI method", meaning a forward incremental chain, where a point is rebuilt from a full and the increments after it. For the newest state, this is not the fastest case.
What does a damaged piece affect?
That depends on the design, and the vendor pages describe the verification, not every consequence.
- Veeam's health check verifies only "the latest restore point in the backup chain", and it may have to read several backup files because the blocks that make up that state "are typically spread out across several backup files" (Health Check for Backup Files). A damaged block that the latest state does not use is outside that check.
- Proxmox recommends re-verifying everything on a schedule: "reverify all backups at least monthly, even if a previous verification was successful", because of "bit rot/data degradation" (Maintenance Tasks).
- In a design where snapshots share chunks, one stored chunk is referenced by many snapshots. Proxmox does not state this directly; it follows that damage to one shared chunk can reach every snapshot that uses it. That is why verification matters more, not less, in such designs.
Space also comes back late. Proxmox pruning "deletes only the metadata, not the underlying backup data"; a later garbage collection removes unused chunks, and chunks newer than the cutoff are not removed. Proxmox sets that cutoff at the oldest running backup, if one exists, or else "24 hours and 5 minutes before the start of the garbage collection".
How can I check what a label means on my system or with my provider?
Ask:
- Where is the one full, how old is it, and when was it last read end to end? In a forever scheme it may be years old.
- What does the store do to keep the chain short, and when does it run? In Veeam, the compact defaults to the last Saturday of the month and the health check to 05:00 on the same day. Both are settings, not fixed behaviour.
- How is damage found? Ask whether the check covers the latest point only, or every point.
- Does the scheme apply to the data type? Bareos limits Always Incremental to file-based backups.
- Is a full restore test done on an old point, not just the newest?
On Veeam, "forever forward incremental" is what you get when "synthetic and active full backups" are both left off. Turning either on gives a forward incremental chain with periodic fulls (Forever Forward Incremental Backup). On Bareos, Accurate = yes is required so that deleted files are not kept in the consolidated backups.
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 & RecoveryRetention that outlasts your ability to restoreHow retention periods, backup chains, keys and obsolete formats come apart, and how to test the path you claim to keep.Read article
- Backup & RecoveryWhat restic and Borg repository checks prove, and what only a restore canWhat restic check, --read-data, borg check and --verify-data actually prove, what they cost, and the restore test that covers what they cannot.Read article
Risks and limits
- Risk: changing the method on a live job, for example turning on periodic fulls, changes the chain layout and the space the repository needs. Check free space first, and read the vendor's own page on switching methods before you change anything. Veeam says its job "does not transform the previously created chain" and instead "creates a new backup chain next to the existing one" (Switching Between Backup Methods). To back out, switch the setting back; that creates another chain next to the existing one, so space is used by both for a time. For a switch between forever forward and forward incremental, the page says the old chain is deleted when the new chain reaches the retention limit.
- Risk: deleting old backup files by hand in any merge-based scheme can break the full that later points depend on. There is no way back unless another copy exists.
- Veeam, Bareos and Proxmox use the same words for different mechanisms. Each description above applies only to that product and version.
- The Bareos page is for bareos-25 (revision 6 November 2026). Proxmox pages show a documentation version (4.2.11-1) but no page date. Veeam pages show update dates.
- Proxmox does not state that damage to a shared chunk reaches every snapshot that uses it; it follows from the design.
- Space figures use decimal gigabytes and round numbers. Real change rates vary by day.
NetLeverage includes a daily full disaster-recovery backup of each hosted server at no extra charge. Talk to an engineer.
Sources
- Forever Forward Incremental Backup, Veeam — Veeam, page updated 2025-08-13, accessed 17 November 2026
- Forever Forward Incremental Backup Retention Policy, Veeam — Veeam, page updated 2025-08-27, accessed 17 November 2026
- Backup Methods, Veeam — Veeam, page updated 2025-08-26, accessed 17 November 2026
- Compact of Full Backup File, Veeam — Veeam, page updated 2025-08-13, accessed 17 November 2026
- Health Check for Backup Files, Veeam — Veeam, page updated 2026-08-04, accessed 17 November 2026
- Switching Between Backup Methods, Veeam — Veeam, page updated 2025-08-12, accessed 17 November 2026
- Synthetic Full Backup, Veeam — Veeam, page updated 2025-07-09, accessed 17 November 2026
- Always Incremental Backup Scheme, Bareos documentation — Bareos project, bareos-25, revision 6 November 2026, accessed 17 November 2026
- Technical Overview, Proxmox Backup Server 4.2.11-1 — Proxmox Server Solutions, date not shown, accessed 17 November 2026
- FAQ, Proxmox Backup Server — Proxmox Server Solutions, copyright 2019-2025, accessed 17 November 2026
- Maintenance Tasks, Proxmox Backup Server — Proxmox Server Solutions, date not shown, accessed 17 November 2026
This article was originally published on .