obsidian-vault/200 projects/220 Web Host Migration/Backup Action Plan — Close the Gaps + Restore.md

created, modified, type, client, project, status, priority, tags, aliases, id

created	modified	type	client	project	status	priority	tags	aliases	id
2026-09-01	2026-09-01	plan	sam	web-host-migration	pending	1	backup restore aws s3 borg kopia restic	backup-action-plan backup-gaps-fix	1848129027-BAP

Backup Action Plan — Close the Gaps + Restore Procedures

Part of Migration Plan — InMotion to Local + AWS Offsite. Follow-up to Backup Architecture — Offsite to AWS S3. Status: PLAN ONLY (no changes yet). Created 2026-09-01 after scope-gap review.

Why this exists

The original offsite S3 plan covered photos + archive + .27 desktop — but review (2026-09-01) found:

1. .13 system configs are not fully in Borg (/home/sam/deployment 23G, voice/audio/pi dirs, etc missing).
2. .13's Borg repo (borg/.13) is NOT in S3 — only .27 is.
3. .35's backup repos (restic 1.1T, kopia 71G on .23) are NOT in S3 and depend on .23 (single point of failure, no backup, on-site).
4. No documented restore procedure exists in the notes.

Facts verified 2026-09-01

Item	Detail
.13 Borg local-to-storage covers	~/Docker, pocketbase-*, dotfiles, .ssh, /mnt/data/family-home-lab, photo-pipeline
.13 NOT in Borg	~/deployment (23G), gst-plugins-rs (2.8G), speech_piper (1.5G), voice_bridge (946M), whisper-api (518M), librespot (842M), voice-agent (82M), mopidy-radionet (3.1M), ~/.agents (844M), ~/.pi (481M), bin, .config
.35 Restic (Backrest) plan sources	/docker-config, /docker-containers, /docker-volumes, trilium-data, /host_fs/mnt/hd
.35 Kopia snapshots	Docker/Containers, trilium-data, var/lib/docker/volumes
.35 repos on .23	restic 1.1T, kopia 71G, borg-repos 462G
.23	NFS file-server; holds ALL .35 repos + .13 borg copies; itself has NO backup
S3 currently	photos/by_date, archive, borg/.27 only

Action plan

Step 1 — Close .13 Borg coverage gap (local, immediate, $0)

• Add to .13 backup.nix local-to-storage paths:
  • ~/deployment
  • ~/gst-plugins-rs, ~/speech_piper, ~/voice_bridge, ~/whisper-api, ~/voice-agent, ~/librespot, ~/mopidy-radionet
  • ~/.agents, ~/.pi, ~/bin, ~/.config
• Exclude: .cache, .npm, .local (junk/regenerable)
• Rebuild + run Borg → verify new archive includes ~/deployment

Step 2 — Add .13's repo + key dirs to S3 (cheap, ~31G)

• Extend offsite.nix S3 seed to include:
  • backup/borg/.13 → borg/.13/ (STANDARD) — tiny (304M)
  • ~/deployment + voice/audio dirs → home/... (DEEP_ARCHIVE) — ~31G ≈ <$0.05/mo
• First seed after Step 1 completes so .13's own config is offsite too

Step 3 — .35 restore path → S3 (DECISION NEEDED on scope)

Given restic repo is 1.1T on .23, options:

• Option A (recommended, restore-focused): Do NOT ship the whole 1.1T repo. Instead:
  • Ship Borg .13 archive copies on .23 (462G of borg-repos) → S3 DEEP_ARCHIVE (these are already dedup'd single-copy restorable archives) — the highest-value offsite target.
  • For .35's restic/kopia, ship ONLY mnt/hd (Immich original photos — already in S3 via photos) — restic/kopia repos remain local (on .23) since .23 will be separately protected.
• Option B (complete): Also ship restic/kopia repos → S3 (~1.1T+) — expensive (first 90 days Standard ≈ $40-80/mo then Deep Archive ≈ $1.6/mo). Only if you want full .35 restore from cloud.
• Decision: pick A or B (recommend A).

Step 4 — Protect .23 (the single point of failure)

• Since .23 is the hub for ALL repo copies: either
  • A: ship .23's borg-repos to S3 (covers Step 3A), or
  • B: give .23 its own backup (rsync to .13? but .13 is on-site too) — offsite is the real fix → S3.
• Recommendation: the S3 copies in Step 2+3A effectively close .23's vulnerability for restore-critical data.

Step 5 — RESTORE PROCEDURES doc (in backup notes)

• Create Restore Runbook (in Backup Architecture — Offsite to AWS S3 or new note): step-by-step for:
  • Restore .27 from borg/.27 (local + S3)
  • Restore .13 from borg/.13 (local + S3) + the /home additions
  • Restore .35 containers/volumes from Backrest/Kopia (local .23) + from S3 if repos offsited
  • Restore Immich/My-Passport from S3 photos (DEEP_ARCHIVE — note hours-to-restore)
  • Restore Gitea (from /Docker/Containers in restic/kopia)
• Add a "test a restore" cadence (e.g. monthly spot-check from S3)

Step 6 — Verify

• After changes: run backup, confirm S3 object counts for new prefixes
• Test-restore ONE path from S3 (prove read path)
• Update monitoring page to show new prefixes

Cost estimate (AUD)

• Step 2 (.13 → S3): +31G → < $0.10/mo steady (DEEP_ARCHIVE)
• Step 3A (borg-repos 462G → S3): $0.01/mo steady DEEP_ARCHIVE after first-90-day Standard ($16/mo transient)
• Total steady-state ≈ $2-4/mo (vs current ~$1-2). Well within budget.

Restore flow summary (target)

1. Machine died, .23 alive: restore .13/.35 from local repos (fast) — documented runbook.
2. House fire: .23 gone → restore .27, .13 configs, photos, Immich from S3 (Glacier hours) — documented runbook.
3. Monthly: spot-test one S3 restore.

---

Linked: Backup Architecture — Offsite to AWS S3, Migration Work Log, Migration Plan — InMotion to Local + AWS Offsite.