Backup — why two layers and not one

Extracted from the master document on 2026-09-04. §N numbering preserved so as not to break cross-references.

37. Why two layers and not one

A single system does not serve two different recovery objectives well:

	Layer 1 — full reconstruction	Layer 2 — granular recovery
Scenario	“the SSD died, start from scratch”	“Vaultwarden got corrupted”, “I deleted a file”
Needs	complete, faithful image of the VM	frequent, lightweight, deduplicated copies
Can be	slow, heavy, infrequent	has to be daily and selective
Tool	Proxmox vzdump	BorgBackup
Frequency	weekly (Sunday 02:00)	daily (03:00)

LAYER 1 — full VM                    LAYER 2 — data and services
Proxmox                              docker-host
  │ vzdump (snapshot + zstd)            │ borg create (dedup + zstd)
  │ weekly                             │ daily
  ▼                                    ▼
VM 100/101 image                     VM 101 /home/borg/repo
  │                                    │ (append-only via restricted SSH)
  │                                    │ prune → compact → mirror
  └──────────► Cloudflare R2 ◄─────────┘
                (rclone crypt, off-site, zero-knowledge)

NEXTCLOUD AIO
  native backup (AIO's own Borg)
        ▼
  /mnt/storage/nextcloud-aio-backup   ← it is one more SOURCE for Layer 2

vzdump → recover the entire VM
Borg   → recover services and data granularly
AIO    → Nextcloud-specific consistency (DB + Redis + files)
R2     → survival beyond the physical node