What is synthetic full backup?
A synthetic full backup is a full backup assembled on backup storage from the last full and the incrementals taken since, rather than by reading all data from production again. Production only supplies a small incremental, and the heavy reading and writing move to the backup storage side.
How is a synthetic full assembled?
- An incremental runs against production and captures only what changed since the last job.
- The backup software reads the previous full and each incremental from backup storage.
- For every block or file it keeps the newest version.
- It writes the result as a new full restore point, which restores on its own like any full.
After the incremental, the production server, its disks and the network to it play no part. The method exists for data sets that have grown until a conventional full overruns the backup window, or where production cannot tolerate another full-volume read.
How much storage I/O does the merge cost?
The cost depends on whether the storage can reference blocks it already holds. With block cloning, the new full is built mostly by pointing at blocks that are already stored, so the merge is largely a metadata operation: little data is read or written, it finishes quickly and the new full takes little extra capacity. Without block cloning, every block of the new full is read from the old chain and written again. That load lands on the backup storage while other jobs and restores may be running, and it can slow them down.
The saving on production is the same in both cases. The difference is whether the backup storage absorbs a full-size copy or avoids it.
Why does corruption carry forward?
A synthetic full never reads production, so it cannot notice that a block in the existing chain is damaged. If an old full or an incremental holds a silently corrupted block, from a bad disk, a software fault or a failed write, the merge treats it as valid and copies or references it into the new full. Each generation inherits the damage, and the restore points look healthy until one is restored.
When is an active full still needed?
An active full reads production directly, so it rebuilds the chain from clean source data. Schedules therefore combine synthetic fulls for most periods with an active full on a longer cycle, or with health checks that read stored blocks and compare them to recorded checksums. Restore tests of the newest synthetic full confirm that the merged result opens in a working system.
Frequently asked questions
Is a synthetic full as good as a regular full?
For restoring, yes. It stands alone like any full. For integrity it is only as good as the chain it was built from.
Does a synthetic full use more storage?
It writes a whole new full unless block cloning lets it share blocks with the old one. The older chain is held until its retention ends, so capacity use rises during the overlap.
Does a synthetic full reduce the backup window?
It shortens the time spent reading production, since only an incremental touches the source. The merge still runs on the storage side.
