What is a backup chain?
A backup chain is a sequence of restore points for one workload, made up of a full backup and the incremental backups that depend on it. Restoring from any point in the chain requires that point and every earlier one back to the full.
What does a backup chain consist of?
A chain starts with a full backup, a complete copy of the data. Each later run stores only the blocks that changed since the previous one. The backup software also keeps metadata describing which blocks belong to which restore point. Together these files or objects form one chain.
| Chain type | How it is built | Restore reads |
|---|---|---|
| Forward incremental | Full, then increments, new full periodically | Full plus increments up to the chosen point |
| Incremental forever | One full, then increments, oldest merged into the full | Full plus current increments |
| Reverse incremental | Newest point kept as a full, older ones as reverse increments | Newest point directly, older ones through rollbacks |
Why does a broken link matter?
Because each increment depends on what came before, a missing or corrupted file in the middle of a chain makes every later restore point unusable. The longer the chain, the more points depend on each early file. Periodic fulls, synthetic fulls and health checks all exist to limit that exposure.
How do backup chains interact with retention?
A restore point cannot be deleted while later points still depend on it. With a 14-day retention and weekly fulls, the oldest full and its increments stay until the whole week that depends on them has expired, so the repository holds more than 14 days of data at times. Retention settings describe the minimum kept, and chain structure decides the actual amount.
What changes on object storage?
On object storage, backup software typically stores data as many small objects and tracks which objects each restore point needs. A new full can then reference existing objects instead of rewriting them, which removes much of the cost of periodic fulls. Immutability also interacts with chains, since locked objects cannot be merged or removed until the lock expires. Backup chains on object storage goes through the details.
Frequently asked questions
How long should a backup chain be?
Long enough to limit the number of fulls and short enough that one damaged file does not affect weeks of restore points. Many policies start a new chain weekly.
Can a single restore point be deleted from the middle of a chain?
Not without affecting the points after it. Backup software removes whole chains or merges the oldest increment into the full.
Does a longer chain make restores slower?
Usually, yes. More increments have to be read and applied, although the effect depends on the repository and the backup format.