backup-patternslisted

# Backup Patterns Review checklist and the choices that are expensive to get wrong. Flag tables for `pg_dump`, `rsync`, and `restic` are in their man pages. ## 3-2-1 Three copies, two media, one offsite. A backup that lives only on the same server as the data is not a backup. ## Choices that matter - **`pg_dump -F c`** (custom) or **`-F d`** (directory) for anything over a few hundred MB. Only these support `-j` parallel restore; `-F p` plain SQL is the slowest both ways. - **`pg_dumpall`** when you need roles and tablespaces, not just one database's contents. - **Pipe rather than stage**: `pg_dump -F c mydb | restic -r <repo> backup --stdin --stdin-filename mydb.dump` avoids a plaintext temp file entirely. - **`rsync --delete`** mirrors exactly - it deletes destination files missing from the source. Dry-run with `-n` before the first real run. - **restic retention** is `forget --keep-daily/--keep-weekly/--keep-monthly --prune`. Without it, storage grows unboundedly. `--dry-run` first. ## Restore testing A backup never tested is not a backup. On a schedule: 1. Restore to a separate location (`/tmp/restore`, a staging server). 2. Verify integrity - row counts, application startup, spot checks. 3. Record the restore duration, so the RTO is known before an incident rather than during one. --- ## Anti-patterns **failure: backup stored only on the same server as the data** A hardware failure, ransomware infection, or accidental `rm -rf` destroys both. The offsite copy i