A smarter forget for restic backup program

Background

If you like/use restic backup program then you probably also know its command restic forget --keep-xyz... which helps cleaning old backups, keeping only the last n ones per hour/day/week/month/year. See appropriate restic documentation for details.

Different approach to forget

This bash-script here provides a different approach: It finds identical consecutive snapshots, then forgets (i.e. deletes/removes) the younger one.

Under the hood

The script uses restic snapshots --json, pipes the (JSON)-result to jq (see https://stedolan.github.io/jq/) and sort (by backup-path and -time) to produces a plain text-listing which contains the tree-hash. When two consecutive snapshots reference the same tree hash then we know that the backup'd data is identical The script finally forgets the second (later, younger) snapshot.

(For safety reasons, the 'forget' command is commented out in the script - so it's a dry-run)

Finally, only snapshots remain which really contains differences to the previous snapshot.

Futher thoughts

• it would be good if restic itself would offer an option to not create a snapshot when no files changed
• you still can combine this script with the provided forget-policies of restic. Propose to run this script first as it removes useless snapshots