What is full backup?
A full backup is a backup run that copies all selected files, volumes or systems into one restore point that can be restored on its own, without any other backup file. Every incremental and differential backup that follows records changes against it, which makes the full the base of the backup chain.
Why is the full the base of a backup chain?
Incremental and differential backups store only changes, so neither restores anything on its own. Each one is a set of differences against an earlier copy, and the chain leads back to a full. If the full is damaged or deleted, every incremental and differential that depends on it becomes unusable, however many there are. The full is the one backup type that stands alone.
How often is a full taken?
Most schedules take a full weekly or monthly, with smaller backups on the days between. Fulls are the slowest and heaviest jobs, since they read all the data and load the source, the network and the target, so they are usually placed in the weekend part of the backup window. A longer gap between fulls reduces that load but lengthens the chain a restore has to read and leaves more restore points dependent on one base.
Where do chains actually fail?
Three failures recur, and none of them shows in a green job status:
- A full expires before its dependents. Retention is set per backup, so a full can reach its end date while incrementals built on it are still kept, leaving restore points that look protected but cannot be rebuilt.
- A chain runs for months with no new full. A single damaged block or missed increment anywhere in a long chain breaks every restore point after it.
- Synthetic fulls overload the backup storage. Building a synthetic full reads the previous full and each increment from the backup storage instead of from production, which moves the heavy read to the target. How fast that runs depends on the storage.
How is the full protected?
A full written to storage with a retention lock is protected from deletion until its lock date. That date has to extend past the last incremental or differential that depends on it, or the dependents lose their base on the day the full is released. A test restore from the chain, not a job log, shows whether the chain still rebuilds.
Frequently asked questions
Does a full backup always read everything?
An active full reads all selected data from the source. A synthetic full builds the same result from existing backup files, without reading production again.
Is a full backup the same as a system image?
No. A full backup is a complete copy of the data selected, which may be files or volumes. A system image is a full copy of a whole machine's disks, including the operating system, used for rebuilds.
Why not take a full every night?
It reads all the data every time, which can exceed the backup window, loads production and stores far more than the changes need. Smaller backups between fulls avoid that.
