homelab-optimized/docs/infrastructure/backup-strategy.md

Backup Strategy

Last updated: 2026-03-21

Overview

The homelab follows a 3-2-1+ backup strategy: 3 copies of data, 2 different storage types, 1 offsite location, plus cloud backup to Backblaze B2.

┌─────────────────────────────────────────────────────────────────────────────┐
│                          BACKUP FLOW                                        │
│                                                                             │
│  Atlantis (Primary) ──── Hyper Backup (weekly) ──── Calypso (Local copy)   │
│       │                                                                     │
│       ├── Syncthing (real-time) ──── Setillo (Tucson, offsite)             │
│       │                                                                     │
│       └── Hyper Backup S3 (weekly) ──── Backblaze B2 (cloud)              │
│                                            │                                │
│  Calypso ──── Hyper Backup S3 (daily) ─────┘                               │
│                                                                             │
│  Guava ──── Restic (daily 3AM) ──── Backblaze B2 (vk-guava, encrypted)     │
│  Jellyfish ──── No backup (risk)                                           │
└─────────────────────────────────────────────────────────────────────────────┘

Backup Tasks

Atlantis → Backblaze B2 (Cloud)

Setting	Value
Task name	Backblaze b2
Schedule	Weekly, Sundays 00:00
Destination	s3.us-west-004.backblazeb2.com
Bucket	vk-atlantis
Encrypted	Yes (client-side)
Versioned	Yes (Smart Recycle)
Rotation	Smart Recycle: keep daily for 7 days, weekly for 4 weeks, monthly for 3 months (max 30 versions)

What's backed up:

• /archive — long-term cold storage
• /documents/msi_uqiyoe — PC sync documents
• /documents/pc_sync_documents — PC sync documents
• /downloads — download staging
• /photo — Synology Photos library
• /homes/vish/Photos — user photo library
• Apps: SynologyPhotos, SynologyDrive, FileStation, HyperBackup, SynoFinder

What's NOT backed up to cloud:

• /volume1/media (~60TB) — too large for cloud backup, replicated to Setillo instead
• /volume1/docker — container data (stateless, can be redeployed from git)

Calypso → Backblaze B2 (Cloud)

Setting	Value
Task name	Backblaze S3
Schedule	Daily, 00:00
Destination	s3.us-west-004.backblazeb2.com
Bucket	vk-concord-1
Encrypted	Yes (client-side)
Versioned	Yes (Smart Recycle)

What's backed up:

• /docker/authentik — SSO provider data (critical)
• /docker/gitea — Git hosting data (critical)
• /docker/headscale — VPN control plane (critical)
• /docker/immich — Photo management DB
• /docker/nginx-proxy-manager — old NPM config (historical)
• /docker/paperlessngx — Document management DB
• /docker/retro_site — Personal website
• /docker/seafile — File storage data
• /data/media/misc — miscellaneous media
• /data/media/music — music library
• /data/media/photos — photo library
• Apps: Gitea, MariaDB10, CloudSync, Authentik, Immich, Paperless, HyperBackup

Atlantis → Calypso (Local Copy)

Setting	Value
Method	Hyper Backup
Schedule	Weekly
Destination	Calypso /volume1/backups/
What	Media, photos, documents
Encrypted	Yes

Atlantis/Calypso → Setillo (Offsite)

Setting	Value
Method	Syncthing (real-time replication)
Destination	Setillo /volume1/syncthing/ (Tucson, AZ)
Distance	~1,000 miles from primary site
What	Docker configs, critical data

Setillo → Backblaze B2 (Cloud)

Setting	Value
Task name	Backblaze B2
Schedule	Scheduled
Destination	s3.us-west-004.backblazeb2.com
Bucket	vk-setillo
Encrypted	No (data encryption disabled — transit only)
Versioned	Yes (Smart Recycle)
Rotation	Smart Recycle: keep daily for 7 days, weekly for 4 weeks, monthly for 3 months (max 30 versions)

What's backed up:

• /backups — backup destination
• /homes/Setillo/Documents — Edgar's documents
• /homes/vish — vish home directory
• /PlexMediaServer/2015_2016_crista_green_iphone_5c — legacy phone photos
• /PlexMediaServer/other — other media
• /PlexMediaServer/photos — photos
• Apps: DownloadStation, FileStation, HyperBackup, SurveillanceStation, SynoFinder, WebDAVServer

Guava (TrueNAS) → Backblaze B2 (Cloud)

Setting	Value
Tool	Restic + Rclone
Schedule	Daily, 03:00 (TrueNAS cron job ID 1)
Destination	s3.us-west-004.backblazeb2.com
Bucket	vk-guava
Repo path	vk-guava/restic
Encrypted	Yes (AES-256, restic client-side encryption)
Password file	/root/.restic-password (chmod 600)
Rclone config	/root/.config/rclone/rclone.conf
Retention	--keep-daily 7 --keep-weekly 4 --keep-monthly 3 --prune

What's backed up:

• /mnt/data/photos (158 GB) — photo library (critical)
• /mnt/data/cocalc (323 MB) — CoCalc notebooks and data
• /mnt/data/medical (14 MB) — medical records (critical)
• /mnt/data/website (58 MB) — website data
• /mnt/data/openproject (13 MB) — project management DB
• /mnt/data/fasten (5 MB) — health data

What's NOT backed up:

• /mnt/data/guava_turquoise (3 TB) — large dataset, not yet assessed
• /mnt/data/jellyfin (203 GB) — media metadata, re-downloadable
• /mnt/data/llama (64 GB) — LLM models, re-downloadable
• /mnt/data/iso (556 MB) — ISOs, re-downloadable

Backup command (manual run):

sudo restic -r rclone:b2:vk-guava/restic \
  --password-file /root/.restic-password \
  backup /mnt/data/photos /mnt/data/cocalc /mnt/data/medical \
         /mnt/data/website /mnt/data/openproject /mnt/data/fasten

Restore command:

sudo restic -r rclone:b2:vk-guava/restic \
  --password-file /root/.restic-password \
  restore latest --target /mnt/data/restore

Check integrity:

sudo restic -r rclone:b2:vk-guava/restic \
  --password-file /root/.restic-password \
  check

Guava Backup → Moon (Browse Access)

The Guava full backup on atlantis is NFS-mounted on moon for browsing:

Setting	Value
Source	atlantis /volume1/archive/guava_full_backup
Mount	moon /home/moon/guava_backup_atlantis
Protocol	NFS v3 over Tailscale (100.83.230.112)
Access	Read-only, moon user (uid 1000)
Persistent	fstab with _netdev,nofail

Disabled Tasks

Task	Host	Reason
Backblaze S3 Atlantis (ID 12)	Atlantis	Old task, replaced by "Backblaze b2" (ID 20)

Hosts Without Backup

Host	Data at Risk	Mitigation
Jellyfish (RPi 5)	1.8TB photos (LUKS2 encrypted NVMe)	LUKS encryption protects at rest, but no redundancy beyond the single drive. Syncthing from phone provides source-of-truth copy.
Homelab VM	Docker data, monitoring databases	Stateless — all compose files in git, data is regenerable. NetBox DB is the main risk
Concord NUC	Home Assistant config, AdGuard	Container data is relatively small and rebuildable

Recommendation: Set up Backblaze B2 backup for Jellyfish (photo archive) — irreplaceable data with no cloud backup. Guava is now covered.

Recovery Procedures

Full NAS Recovery (Atlantis)

1. Replace failed hardware / reinstall DSM
2. Restore from Calypso (fastest — local, weekly copy)
3. Or restore from Backblaze B2 (slower — download over internet)
4. Redeploy Docker stacks from git (all GitOps-managed)

Service Recovery (Any Host)

1. All Docker stacks are in git (hosts/ directory)
2. Portainer GitOps auto-deploys on push
3. Just create the Portainer stack pointing to the compose file
4. Service-specific data may need restore from backup

Critical Service Priority

Priority	Service	Backup Source	Recovery Time
1	Authentik (SSO)	Calypso B2 daily	~30 min
2	Gitea (Git)	Calypso B2 daily	~30 min
3	NPM (Reverse Proxy)	Calypso B2 daily / matrix-ubuntu local	~5 min (redeploy)
4	Plex (Media)	Atlantis B2 weekly	~1 hr (metadata only, media on disk)
5	Paperless (Documents)	Calypso B2 daily	~30 min

Monitoring

• DIUN: Monitors container image updates (weekly, ntfy notification)
• Uptime Kuma: Monitors service availability (97 monitors)
• HyperBackup: Sends DSM notification on backup success/failure
• Backblaze B2: Dashboard at https://secure.backblaze.com/b2_buckets.htm

Related Documentation

• Storage Topology — detailed storage layout per host
• Image Update Guide — how services are updated
• Offline & Remote Access — accessing services when internet is down
• Ansible Playbook Guide — backup_configs.yml and backup_databases.yml playbooks