Files
felhom.eu/documentation/audits/night-fixes-2026-10-05/partA/lab-real-restic-vs-simulation.txt
T

18 lines
1.0 KiB
Plaintext

R-867 lab, 2026-10-05, on DooPlex scratchpad (local repo, deleted after). restic binary copied out of
gitea.dooplex.hu/admin/felhom-controller:0.293.0: `restic 0.14.0 compiled with go1.19.8 on linux/amd64`.
Snapshot set: 2 groups (host box, tags opengist / bookstack), one snapshot per night at 02:15:05 UTC from
2026-08-20 to 2026-10-05, box OFF 2026-09-15..17 (no snapshots), a same-day manual run at 15:00 on 2026-09-10
and 2026-09-30. 92 snapshots, all UTC.
Real policy: restic forget --group-by host,tags --keep-daily 7 --keep-weekly 4 --keep-monthly 6 --dry-run --json
-> removes 72 snapshots
Test simulation restic0140Plan (offbox_window_test.go) over the same `restic snapshots --json`:
-> removes 72 snapshots
diff of the two sorted id lists: EMPTY (identical)
The guard (v0.294.0 candidate) on that plan at now = 2026-10-05T02:20:00Z, maxRemove 500:
GUARD removes=70 refusal=""
(the 2 kept back are the 2026-09-30 02:15 copies superseded by the same-day 15:00 run — R-824's exclusion;
removed in a later window once 7 calendar days old)