boma/docs/decisions/022-backup.md

# ADR-022 — Backup & disaster recovery: data-only restic, off-cluster pull node, 3-2-1

## Status

Accepted (2026-06-10). Resolves TODO 3.8 ("ensure the right things are backed up,
incl. DB dumps") and `CAPABILITIES.md` §9 (backup engine / off-site / air-gap, all
"planned"). Grounds ADR-011's "backup-first" and "snapshot + backup" language, which
assumed a backup policy existed but never defined one.

**Doctrine ADR.** It pins the recovery model, backup engine, topology, per-service
contract, encryption/escrow, restore-testing tiers, retention, alerting, and USB
air-gap mechanism. It does **not** build any of them — the `backup` role, `fisi`
node, per-service `backup__*` declarations, and `BACKUP.md` files do not exist yet.
Designed now, built in the implementation plan referenced at the foot of this ADR.

## Context

boma has no defined backup policy. The ADRs assume one exists — ADR-011 makes
"backup-first" the rule for stateful upgrades and "snapshot + backup" the rollback
path — but nothing specifies *what* gets backed up, *how* it stays consistent, *where*
copies live, *how* they are encrypted, or *whether restores actually work*.
`CAPABILITIES.md` §9 sketches an intent (PBS + restic, pCloud off-site, USB air-gap)
but commits to nothing.

The gap is not just theoretical. Every boma service is stateful in some dimension:
DB contents, bind-mount data dirs, the Vaultwarden vault that holds every secret in
the stack. Without a backup policy the IaC is not reproducible from nothing; it is
reproducible-modulo-data. This ADR closes that gap.

## Decision

### 1. Recovery model — data-only backups, rebuild from code (Model A)

boma's *configuration* is reproducible from this repo: Terraform recreates the VM,
Ansible re-renders the Docker Compose stack. Backups therefore protect **state only** —
DB contents, bind-mount data dirs, Vaultwarden's vault — not whole-VM images.

Recovery sequence: Terraform re-provisions the VM → Ansible redeploys → restic
restores the data. **No Proxmox Backup Server (PBS) in v1.** This keeps the 3-2-1
topology cheap, fits pCloud's 1 TB comfortably, and turns every restore drill into
a continuous proof that the IaC *and* the backups both work.

Trade-off accepted: recovery is slower than a VM-image restore (a full Ansible run
plus data restore, potentially hours), and it bets the repo is complete enough to
rebuild from nothing — which Tier-2 restore testing (Decision 8) exists to verify.
**PBS (Model B) or a per-host hybrid (Model C) can be added later** if real-world RTO
proves too slow; nothing here precludes it.

### 2. One backup tier, ~24 h RPO

A single tier: nightly backup of all state, accepting up to ~24 h of data loss across
the board. No per-data-type tiering yet — revisit once there is real-world data and
experience to justify the added machinery.

### 3. Engine — restic (data) + rclone (off-site); no second encryption layer

- **restic** captures state into an encrypted, deduplicated repository.
- **rclone** replicates the repo to pCloud (pCloud has no good headless Linux client;
  rclone has a first-class pCloud backend).
- restic encrypts the repo at rest, so rclone copies **ciphertext only** — no second
  encryption layer, no pCloud "crypto folder."

No PBS in v1 (see Decision 1).

### 4. Topology — central pull node (`fisi`), off the cluster; `backup_hosts` group

A single backup node owns the canonical restic repo. It is **off the Proxmox cluster**
— an independent failure domain, so copy 2 survives a PVE node (or the whole cluster)
dying. This mirrors the existing pattern for `ubongo` (control) and `askari`
(off-site): a manually-provisioned physical node in its own inventory group, still
Ansible-managed (the `base` role applies, plus a `backup` role).

**Pull model.** `fisi` holds SSH keys to each host; per service it runs the declared
dump command remotely, pulls the declared paths read-only, then `restic` snapshots the
staged data into its local repo. **Hosts hold no backup credentials and cannot reach
the repo** — a compromised or ransomwared service host cannot delete backup history.

**Node assignment:** `fisi` (an HP Elite 600 G9 tower) is penciled in / provisional —
the *role* ("the backup node") is load-bearing; the physical assignment may be
revisited when all hardware is on hand. `fisi` holds **2× 8 TB HDDs in a mirror**
(ZFS or mdraid → 8 TB usable, survives one disk failure). It owns the repo, runs the
pull orchestration, runs `rclone → pCloud`, and docks the USB air-gap drives
(Decision 11).

**Inventory:** a new `backup_hosts` group is added to both inventories, structured
like `control` and `offsite_hosts`. The `base` role applies.

### 5. 3-2-1 mapping

| Copy | Location | Medium | Off-site? | Notes |
|---|---|---|---|---|
| 1 | Live data on each host | NVMe/SSD | no | The working data |
| 2 | `fisi` restic repo | 8 TB HDD mirror | no (on-site, off-cluster) | Canonical repo |
| 3 | pCloud (via rclone) | Cloud | **yes** | Encrypted ciphertext; **sync-coupled** (see Consequences) |
| +4 | USB air-gap drive(s) | Removable HDD, **offline** | yes (stored off-site) | The **immutable backstop**; rotated |

≥3 copies, ≥2 media, ≥1 off-site — 3-2-1 satisfied, with the air-gap drive as a
fourth, offline copy that no online compromise can reach.

### 6. Per-service backup contract — `backup__*` data + `BACKUP.md`; governance

Each service role declares its backup needs in role vars — the same render-from-data
pattern boma uses for `access__*`/`ACCESS.md` (ADR-021):

```yaml
backup__service: nextcloud   # identifier; matches the role / compose project
backup__state: true          # false = stateless → no BACKUP.md (pair with a reason)
backup__paths:               # bind-mount dirs / files holding state ([] = none)
  - /srv/nextcloud/data
backup__dumps:               # logical app-consistent dumps ([] = none)
  - cmd: "docker compose -p nextcloud exec -T db pg_dump -U {{ vault.nextcloud.db_user }} nextcloud"
    dest: nextcloud-db.sql
backup__quiesce: false       # true = stop→back up→restart escape hatch (Decision 7 B)
```

The pull orchestrator reads these (rendered from inventory) and, per service: SSH in →
run the dumps → pull the dump files + declared paths read-only → `restic` snapshot. A
service with **no** `backup__paths` must explicitly declare `backup__state: false` with
a reason; omission is never an implicit "nothing to back up." (`backup__state` and the
list-form `backup__dumps` are this ADR's resolution of the spec's open "declared, not
silent" point.)

**`BACKUP.md` becomes a required per-service doc** alongside `SECURITY.md`,
`VERIFY.md`, and `ACCESS.md`, **rendered from the role's `backup__*` data**, documenting:
what state exists, what is backed up, the dump command, and the per-service restore
procedure. A template lives at `docs/backup/service-backup-template.md`. A **stateless**
service declares `backup__state: false` (with a reason) in its role vars and gets **no**
`BACKUP.md`.

**Governance — runbook + gate, not scaffold (consistent with ADR-021).** Three light
touches mirror how `SECURITY.md`, `VERIFY.md`, and `ACCESS.md` are enforced: the
service checklist (`docs/security/service-checklist.md`) gains a backup item; the
`new-role` runbook gains a fill/render/`check-backup` step (copy
`docs/backup/service-backup-template.md` into `roles/<service>/BACKUP.md` and
populate the `backup__*` data); and a checklist gate blocks service clearance until
the record exists and a restore drill confirms it (or a deviation is recorded in
`accepted-risks.md`). The dormant `/check-backup` verifier is the automated check
analogue of `/check-access` (ADR-021). **No automated lint script gates `BACKUP.md`
presence** — same manual-copy-plus-review pattern the sibling records use. The design
document's "make lint gates its presence" wording is superseded by this governance
choice.

### 7. Consistency — logical dumps first; quiesce as escape hatch

- **Default:** databases are captured with logical dumps (`pg_dump` / `mysqldump`) —
  portable, version-independent, restorable to a fresh DB. Plain data dirs are backed
  up as files. No downtime required.
- **Escape hatch:** a service whose data cannot be dumped live declares a quiesce step
  (stop container → back up volume → restart) via `backup__quiesce` in the same contract.
- ZFS/filesystem snapshots are **not** used as the sole DB method (only
  crash-consistent for a live database).

This is agnostic to the open central-vs-per-app database question (TODO 3.9): either
way, each service declares how to dump its own data.

### 8. Restore testing — two tiers; `ubongo` stays bare Debian

- **Tier 1 — weekly, automated, rolling restore-verify.** Pick the next service in
  rotation, restore its latest snapshot into a throwaway container on `ubongo`
  (reusing the Molecule harness, ADR-015), start the app against the restored data,
  and run that service's `VERIFY.md` checks (ADR-008/017). This catches the failure
  that actually kills people — *silently corrupt or unrestorable backups*. Failures
  alert via ntfy.
- **Tier 2 — semi-annual full DR rehearsal,** driven from `ubongo` onto PVE staging.
  Rebuild a host from zero via Terraform + Ansible + restic restore on the staging
  cluster. This validates the whole Model-A recovery chain. **At least once a year the
  rehearsal exercises the paper-secret break-glass path** (Decision 10) end-to-end.

**`ubongo` stays bare Debian, not a hypervisor (ADR-015 unchanged).** Its role is to
be the independent recovery anchor — "the tool used to rebuild the cluster must not
live inside the thing it rebuilds." Higher-fidelity real-VM testing is better served
by the PVE staging environment (same hardware class, same cluster, same provisioning
path). `ubongo`'s 1 TB NVMe gives ample room for Tier-1 dataset restores; disk
headroom (not CPU/RAM) is the first thing to watch as data grows (`/capacity-review`).

### 9. Retention — GFS via restic

Starting policy: `--keep-daily 7 --keep-weekly 4 --keep-monthly 6 --keep-yearly 1`.
`restic forget --prune` runs nightly on `fisi`'s repo; pCloud mirrors the pruned repo.
Tune once real repo growth is observed.

### 10. Encryption + key escrow + break-glass

restic encrypts the repo at rest, so **one secret — the restic repo password —
protects all copies uniformly** (`fisi`, pCloud, USB). One thing to escrow, not three.

**Escrow locations:**
- **`fisi`, root-only** (plus in the Ansible vault) — so backups run non-interactively
  and `fisi` is redeployable.
- **Vaultwarden** — the day-to-day human-accessible copy.
- **Paper, in a physical safe (off-site)** — the break-glass root of trust; the only
  copy that survives "everything is down."

**The paper holds *two* secrets:** (1) the **restic repo password** (to read any
backup at all) and (2) the **Ansible vault master password** (to rebuild hosts from
the repo — normally from Vaultwarden via `rbw`, which is itself down in a from-zero
recovery). With both on paper, the break-glass chain has **no circular dependency**:
paper → restic restores Vaultwarden + repo data → the vault password (from paper)
drives Terraform/Ansible re-provisioning → services return, `rbw` works again.

**`mamba` (laptop) is the break-glass clone** (ADR-015): repo + toolchain + mesh +
`rbw`, with Terraform state synced to it — the rebuild can be driven from `mamba` if
`ubongo` is also gone. The paper sheet doubles as a short break-glass runbook assuming
zero running boma infrastructure: install restic on any machine, point it at pCloud
*or* a USB drive with the password, restore Vaultwarden first, then rebuild with the
vault password.

### 11. USB air-gap — plug-and-go cold copy

A **udev rule on `fisi` matching an allowlist of known drive serials** triggers a
systemd unit / script that: mounts the drive, confirms it is an expected drive, runs
**`restic copy` from the local repo → a restic repo on the USB drive** (same
password → ciphertext if lost/stolen), runs `restic check` on the USB copy, unmounts,
and **notifies via ntfy** with the result. Only allowlisted serials trigger anything —
a rogue USB does nothing.

`restic copy` (not rsync) so the USB is itself a valid restic repo, restorable
directly in a break-glass with nothing else alive. Drives are rotated and **stored
off-site** — a second geographic off-site copy independent of pCloud.

### 12. Failure alerting — guard against silent death

Success/failure pings alone miss the worst case (*the job silently stopped running*):

- **Dead-man's-switch:** every successful nightly run pings an **Uptime Kuma push
  monitor**; no ping in ~25 h → alert.
- **Immediate failure → ntfy** on any job or dump-step error.
- **Weekly `restic check`** for repo integrity → alert on corruption.
- **Tier-1 restore-verify failures → ntfy.**
- *(Later)* emit last-success timestamp + repo size as Prometheus metrics for a
  Grafana panel (fits ADR-018's monitoring direction; not required for v1).

### 13. Schedule

- **Nightly backup run (~02:00–04:00),** driven by `fisi` (pull): per host →
  run dumps → pull paths read-only → `restic` snapshot → `restic forget --prune`
  → `rclone sync` → pCloud. Sequential, off-hours.
- **Tier-1 restore-verify:** weekly, rolling one service per run, on `ubongo`.
- **Tier-2 DR rehearsal:** semi-annual on staging; ≥1/year exercises the paper path.
- **USB air-gap:** manual, approximately monthly, whenever a drive is docked.

## Consequences

- boma now has a defined, end-to-end backup policy that closes the gap ADR-011 left
  open; "backup-first" and "snapshot + backup" are no longer assumed.
- Every service role that holds state must declare its backup contract (`backup__*`
  vars + `BACKUP.md`); stateless services declare `backup__state: false`. Cost:
  per-service declarations and a rendered doc to maintain (mitigated by the new-role
  runbook step + checklist gate).
- **pCloud is off-site but sync-coupled** — `rclone sync` propagates deletions (a
  prune, or a malicious wipe of `fisi`'s repo, replicates to pCloud). The **USB
  air-gap drive is the only truly immutable copy**; pCloud's own file-version history
  is enabled as a secondary cushion.
- **`fisi` is the crown-jewel host** — it holds an encrypted copy of all state, so it
  receives full `base` hardening and tight access. restic encryption means a stolen
  `fisi`, USB drive, or pCloud blob yields ciphertext only.
- **pCloud's 1 TB is the off-site capacity ceiling.** Data-only backups fit for years
  at homelab scale; flag for `/capacity-review` if the repo trends toward ~1 TB.
- Recovery time under Model A (full Ansible run + data restore) is potentially hours —
  slower than a VM-image restore. PBS/Model B is explicitly deferred, not rejected.
- The paper break-glass must be kept current (restic password + vault password). An
  outdated paper sheet is the one failure mode this ADR cannot prevent mechanically —
  the semi-annual DR rehearsal is the human control.

Full design rationale and worked examples: `docs/superpowers/specs/2026-06-10-backup-strategy-design.md`.
Build path (roles, topology, tests): `docs/superpowers/plans/2026-06-10-backup-strategy.md`.

## Related

ADR-002 (security baseline: hardening applied to `fisi`), ADR-004 (one service = one
role; per-service doc conventions), ADR-008 (testing methodology; Molecule harness
reused for Tier-1), ADR-011 (update management: backup-first rule now grounded),
ADR-015 (`ubongo` recovery model; `mamba` break-glass clone; bare-Debian invariant),
ADR-017 (`VERIFY.md` checks reused in Tier-1 restore-verify), ADR-018 (logging/Alloy
→ ntfy alerting path), ADR-019 (Proxmox tags; `backup_hosts` group), ADR-021
(render-from-data pattern: `access__*`/`ACCESS.md` → `backup__*`/`BACKUP.md`;
runbook+gate governance model).