Files
felhom-controller/controller/internal/backup/offbox_restore.go
T
admin 062357f778 v0.148.0 — coherent snapshot pairs + an offsite restore that actually restores (R-43 + R-44)
Closes the two findings from DIAG-immich-restore-2026-07-19. Viktor deleted 11
immich photos to test offsite restore; both runs flashed success and the photos
stayed gone. Two independent defects.

R-43 — no offsite path could restore a database. All three buttons were
file-only: the two "visszaállítás" actions staged to a scratch folder and never
touched postgres, and place-to-live merged only MISSING files. For a DB-indexed
app the bytes returned and the app still could not see them. The dump was
carried INTO every snapshot and could never be replayed OUT of one.

New ReconstituteFromOffsite (/backup/offbox/reconstitute): safety dump → stop →
files overwritten to the snapshot version → start → the snapshot's own dump
replayed → health wait. Two invariants:
  - nothing is ever deleted (-a, no --ignore-existing, no --delete): a file
    created after the snapshot survives as an extra;
  - the undo exists before the act — the pre-restore- dump is verified ON DISK
    before anything is stopped, overwritten or replayed; if it cannot be taken
    the operation refuses with zero changes.
The replay reads the SCRATCH unit: the live unit is never overwritten, so
replaying from it would replay the current DB over itself and restore nothing.

R-44 — a manual push shipped an unrefreshed dump (up to ~24h old). That day's
predated the customer's account by four hours and probed to asset:0/user:0/
album:0 inside 52MB whose bulk was immich's shipped geodata. Every run, manual
AND nightly, now refreshes dumps + units BEFORE capturing. Order is the
mechanism: the gap can only ADD files the DB does not reference yet, never
remove one it does. Manifests carry offsite_run_id + dumps_at, so coherence is
verifiable at restore time rather than assumed; the periodic refresh carries a
prior stamp forward and never invents one.

Honesty surfaces, all warn-level and none a gate: unstamped (pre-v0.148) pairs
report their skew, ValidateDump gained an EXACT-match accounts-table sniff for
customer-empty dumps, the completion flash states an outcome instead of a
mechanism, and the missing-only button now says what it does NOT do.

11 tests; 5 red-proofs run and reverted. Two of those found real test weaknesses
rather than confirming strength — the first undo mutation was caught by a second
guard, and the first table-matching test did not discriminate between the two
matchers at all. Both tests were rewritten to the cases that separate them.

NOT in scope: R-41's catalog invariant check, nightly cadence, retention, quota
math, tier-2, and v0.147.x progress semantics beyond one added phase line.

Live acceptance (§9) has NOT run: no capability-map flip, customer-restore row
stays MISSING, R-3 stays DRAFT.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01P9Nn14TWGzKoqAJAiVwC2s
2026-07-19 12:21:16 +02:00

433 lines
18 KiB
Go
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
package backup
import (
"context"
"encoding/json"
"fmt"
"os"
"os/exec"
"path/filepath"
"strings"
"time"
)
// Offsite restore rework (Task 3a §7). With mandatory userdata now in snapshots, restore needs three
// changes over the old dump-to-rootfs-scratch:
// 1. scratch relocated off the ~8 GB guest rootfs onto a data drive, behind a headroom gate (F-A1);
// 2. a unit-only DEFAULT restore (`--include <absolute-unit-path>`, SP-3.2) — full is a deliberate,
// size-gated second action;
// 3. place-to-live = a missing-only merge (never --delete) so the SQ3 immich case is restorable
// from offsite alone.
// ID-first everywhere (§3): `restic stats --tag` is UNPROVEN on 0.14.0, so the size lookup resolves the
// snapshot ID via `snapshots latest --tag` and calls `stats <ID>`.
const (
// offboxUnitOnlyFreeFloor — a unit-only restore needs at least this much free on the scratch drive.
// Catalog recovery units are MB1 GB (SQ4); 2 GiB is a safe floor without a per-snapshot size probe.
offboxUnitOnlyFreeFloor = int64(2) << 30
)
// SetOffboxFreeFn overrides the restore free-space probe (tests; the Windows go-test host has no df).
func (m *Manager) SetOffboxFreeFn(fn func(path string) int64) { m.offboxFreeFn = fn }
// SetOffboxFullPlaceCopier overrides the FULL-restore overwrite copier (tests; no rsync needed).
func (m *Manager) SetOffboxFullPlaceCopier(fn func(src, dst string) (int, error)) {
m.offboxFullPlaceCopier = fn
}
// SetSafetyDumpFn overrides the pre-restore safety dump (tests; no Docker needed).
func (m *Manager) SetSafetyDumpFn(fn func(ctx context.Context, db DiscoveredDB, dumpDir string) DumpResult) {
m.safetyDumpFn = fn
}
// offboxFree returns the free-space probe (nil seam → the real diskFreeBytes).
func (m *Manager) offboxFree() func(string) int64 {
if m.offboxFreeFn != nil {
return m.offboxFreeFn
}
return diskFreeBytes
}
// diskFreeBytes returns available bytes on the filesystem holding path (0 on any error). Mirrors
// appexport.DiskFree; kept local so the backup package needs no cross-package dependency.
func diskFreeBytes(path string) int64 {
ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
defer cancel()
out, err := exec.CommandContext(ctx, "df", "--output=avail", "-B1", path).Output()
if err != nil {
return 0
}
lines := strings.Split(strings.TrimSpace(string(out)), "\n")
if len(lines) < 2 {
return 0
}
var size int64
fmt.Sscanf(strings.TrimSpace(lines[1]), "%d", &size)
return size
}
// offboxUnitPathOf returns the snapshot path that is the recovery unit for stack (suffix
// backups/primary/<stack>), or "" if none is present.
func offboxUnitPathOf(paths []string, stack string) string {
suffix := "/backups/primary/" + stack
for _, p := range paths {
if strings.HasSuffix(p, suffix) {
return p
}
}
return ""
}
// offboxLatestSnapshot resolves the newest snapshot for stack: its short ID + captured paths, via
// `snapshots latest --tag <stack> --json`. When the tag spans more than one group (old unit-only shape
// + new enlarged shape), it returns the newest by time.
func (m *Manager) offboxLatestSnapshot(ctx context.Context, stack string) (id string, paths []string, err error) {
t := m.settings.GetOffboxTarget()
base, env := m.offboxBaseArgs(t)
sctx, cancel := context.WithTimeout(ctx, offboxProbeTimeout)
defer cancel()
out, serr := m.runner()(sctx, env, append(append([]string{}, base...), "snapshots", "latest", "--tag", stack, "--json")...)
if serr != nil {
return "", nil, fmt.Errorf("offbox snapshots %s: %w: %s", stack, serr, truncate(out))
}
var snaps []struct {
ShortID string `json:"short_id"`
ID string `json:"id"`
Time time.Time `json:"time"`
Paths []string `json:"paths"`
}
if json.Unmarshal(out, &snaps) != nil || len(snaps) == 0 {
return "", nil, fmt.Errorf("offbox: nincs pillanatkép a(z) %s alkalmazáshoz", stack)
}
best := 0
for i := 1; i < len(snaps); i++ {
if snaps[i].Time.After(snaps[best].Time) {
best = i
}
}
id = snaps[best].ShortID
if id == "" {
id = snaps[best].ID
}
return id, snaps[best].Paths, nil
}
// offboxSnapshotSize returns the restore-size (logical bytes) of ONE snapshot via `stats <ID> --json`
// (default mode — for a single snapshot ID this is exactly that snapshot's on-disk-when-restored size,
// the correct headroom meaning; SP-1). ID-first: never `stats --tag` (unproven on 0.14.0).
func (m *Manager) offboxSnapshotSize(ctx context.Context, id string) (int64, error) {
t := m.settings.GetOffboxTarget()
base, env := m.offboxBaseArgs(t)
sctx, cancel := context.WithTimeout(ctx, offboxProbeTimeout)
defer cancel()
out, err := m.runner()(sctx, env, append(append([]string{}, base...), "stats", id, "--json")...)
if err != nil {
return 0, fmt.Errorf("offbox stats %s: %w: %s", id, err, truncate(out))
}
var st struct {
TotalSize int64 `json:"total_size"`
}
if json.Unmarshal(out, &st) != nil || st.TotalSize <= 0 {
return 0, fmt.Errorf("offbox: a(z) %s pillanatkép mérete ismeretlen", id)
}
return st.TotalSize, nil
}
// offboxRestoreScratchDir returns the on-DATA-DRIVE scratch dir for an app's offsite restore
// (<nsRoot>/backups/offsite-restore/<app>) plus the namespace root (an existing dir, for the free-space
// probe). NEVER cfg.Paths.DataDir (the rootfs — the F-A1 filler). App's HDD drive first; else the first
// schedulable storage path; else a Hungarian refusal.
func (m *Manager) offboxRestoreScratchDir(stack string) (scratch, nsRoot string, err error) {
// offsiteRestoreRootFor is THE place `backups/offsite-restore` is spelled (offbox_verify_copies.go)
// — the listing/delete surface must resolve byte-identical paths to the ones written here.
scratchFor := func(root string) (string, string) {
return filepath.Join(m.offsiteRestoreRootFor(root), stack), m.namespaceRoot(root)
}
isNet := func(path string) bool { return m.settings != nil && m.settings.IsNetworkStoragePath(path) }
// (1) the app's own drive — preferred, but ONLY if it is not NETWORK storage (F-3afix-1). restic
// restores uid/gid/setgid fully onto a LOCAL fs (SP-3.3); a squashed network scratch would feed
// PlaceOffsiteRestore wrong-owner files — the F-6C-1 silently-broken-restore class, offsite-side.
if m.stackProvider != nil {
if hdd := strings.TrimSpace(m.stackProvider.GetStackHDDPath(stack)); hdd != "" && !isNet(hdd) {
s, nr := scratchFor(hdd)
return s, nr, nil
}
}
// (2) the first NON-network schedulable path.
if m.settings != nil {
for _, sp := range m.settings.GetSchedulableStoragePaths() {
if strings.TrimSpace(sp.Path) != "" && !sp.IsNetwork() {
s, nr := scratchFor(sp.Path)
return s, nr, nil
}
}
// (3) last resort ONLY: any schedulable path, with a loud WARN — a network scratch cannot
// guarantee ownership fidelity under root_squash.
for _, sp := range m.settings.GetSchedulableStoragePaths() {
if strings.TrimSpace(sp.Path) != "" {
m.logger.Printf("[WARN] [offbox] %s: restore scratch on network storage %s — ownership fidelity not guaranteed under squash", stack, sp.Path)
s, nr := scratchFor(sp.Path)
return s, nr, nil
}
}
}
return "", "", fmt.Errorf("nincs elérhető adatmeghajtó a visszaállításhoz")
}
// RestoreOffboxScratch restores an app's latest offsite snapshot to an on-data-drive scratch dir
// (non-destructive — never overwrites live data). full=false (the default) restores the recovery UNIT
// only (`--include <absolute-unit-path>`, SP-3.2); full=true restores the whole snapshot (unit +
// mandatory userdata) behind a size×1.1 headroom gate. Fail-closed: an unknown snapshot size refuses a
// full restore.
func (m *Manager) RestoreOffboxScratch(ctx context.Context, stack string, full bool) error {
if !m.OffboxConfigured() {
return fmt.Errorf("off-box backup not configured")
}
if !isSafeStackName(stack) {
return fmt.Errorf("invalid stack name")
}
id, paths, err := m.offboxLatestSnapshot(ctx, stack)
if err != nil {
return err
}
unitPath := offboxUnitPathOf(paths, stack)
if unitPath == "" {
return fmt.Errorf("a(z) %s pillanatképében nincs mentési egység — a visszaállítás nem indítható", stack)
}
scratch, nsRoot, err := m.offboxRestoreScratchDir(stack)
if err != nil {
return err
}
// Headroom gate (F-A1) — probed on the namespace root (an existing dir).
free := m.offboxFree()(nsRoot)
if full {
size, serr := m.offboxSnapshotSize(ctx, id)
if serr != nil {
// SizeUnknown never renders as fits — fail closed.
return fmt.Errorf("A mentés mérete nem állapítható meg — a teljes visszaállítás biztonsági okból nem indítható.")
}
need := size + size/10 // ×1.1
if free < need {
return fmt.Errorf("Nincs elég szabad hely a visszaállításhoz (%s szükséges, %s szabad).", humanizeBytes(need), humanizeBytes(free))
}
} else if free < offboxUnitOnlyFreeFloor {
return fmt.Errorf("Nincs elég szabad hely a visszaállításhoz (%s szükséges, %s szabad).", humanizeBytes(offboxUnitOnlyFreeFloor), humanizeBytes(free))
}
// F-A1 hygiene: drop the legacy rootfs scratch (DataDir/offbox-restore/<app>) best-effort.
legacy := filepath.Join(m.cfg.Paths.DataDir, "offbox-restore", stack)
if _, sErr := os.Stat(legacy); sErr == nil {
if rmErr := os.RemoveAll(legacy); rmErr != nil {
m.logger.Printf("[WARN] [offbox] could not remove legacy rootfs restore scratch %s: %v", legacy, rmErr)
} else {
m.logger.Printf("[INFO] [offbox] removed legacy rootfs restore scratch %s", legacy)
}
}
if err := os.MkdirAll(scratch, 0o755); err != nil {
return fmt.Errorf("restore dir: %w", err)
}
t := m.settings.GetOffboxTarget()
base, env := m.offboxBaseArgs(t)
rctx, cancel := context.WithTimeout(ctx, offboxBackupTimeout)
defer cancel()
m.unlockStale(rctx, base, env) // pre-restore hygiene
args := []string{"restore", id, "--target", scratch}
if !full {
args = append(args, "--include", unitPath) // SP-3.2: absolute snapshot unit path = unit-only
}
out, rerr := m.resticStep(rctx, env, base, "restore:"+stack, args...)
if rerr != nil {
return fmt.Errorf("offbox restore %s: %w: %s", stack, rerr, truncate(out))
}
m.logger.Printf("[INFO] [offbox] restored %s (%s, full=%v) → %s", stack, id, full, scratch)
return nil
}
// OffboxRestorePrepareFull resolves the latest snapshot's restore-size and verifies scratch headroom
// for a FULL restore WITHOUT starting it (the two-step size-first gate). Returns the human size on
// success, or a Hungarian error to flash on refusal (size unknown / no headroom — fail-closed).
func (m *Manager) OffboxRestorePrepareFull(ctx context.Context, stack string) (string, error) {
if !m.OffboxConfigured() {
return "", fmt.Errorf("off-box backup not configured")
}
if !isSafeStackName(stack) {
return "", fmt.Errorf("invalid stack name")
}
id, _, err := m.offboxLatestSnapshot(ctx, stack)
if err != nil {
return "", err
}
size, serr := m.offboxSnapshotSize(ctx, id)
if serr != nil {
return "", fmt.Errorf("A mentés mérete nem állapítható meg — a teljes visszaállítás biztonsági okból nem indítható.")
}
_, nsRoot, derr := m.offboxRestoreScratchDir(stack)
if derr != nil {
return "", derr
}
need := size + size/10
if free := m.offboxFree()(nsRoot); free < need {
return "", fmt.Errorf("Nincs elég szabad hely a visszaállításhoz (%s szükséges, %s szabad).", humanizeBytes(need), humanizeBytes(free))
}
return humanizeBytes(size), nil
}
// OffboxFullScratchReady reports whether a (non-empty) full-restore scratch exists for stack — the gate
// for showing the place-to-live action. PlaceOffsiteRestore re-validates per-path completeness.
func (m *Manager) OffboxFullScratchReady(stack string) bool {
if !isSafeStackName(stack) {
return false
}
scratch, _, err := m.offboxRestoreScratchDir(stack)
if err != nil {
return false
}
if fi, sErr := os.Stat(scratch); sErr != nil || !fi.IsDir() {
return false
}
entries, _ := os.ReadDir(scratch)
return len(entries) > 0
}
// placement is one source→dest pair for place-to-live: src is the reconstructed absolute path under the
// scratch (SP-3.1), dst is the live location under the app's current namespace root.
type placement struct {
src string
dst string
isUnit bool
}
// mapOffsiteRestorePaths maps a completed full-scratch restore to live placements (pure). The anchor
// oldNs is derived by trimming backups/primary/<stack> off the unit path (the snapshot may come from a
// DIFFERENT drive after churn — liveNsRoot is where it goes). Refuses the WHOLE placement (no partial
// writes) on: no unit path; a path outside oldNs (escape); a `..` segment; a non-unit path in the
// reserved backups/ zone.
func mapOffsiteRestorePaths(snapPaths []string, stack, scratch, liveNsRoot string) ([]placement, error) {
unitSuffix := "/backups/primary/" + stack
oldNs := ""
for _, p := range snapPaths {
if strings.HasSuffix(p, unitSuffix) {
oldNs = strings.TrimSuffix(p, unitSuffix)
break
}
}
if oldNs == "" {
return nil, fmt.Errorf("a pillanatképben nincs mentési egység (backups/primary/%s)", stack)
}
out := make([]placement, 0, len(snapPaths))
for _, p := range snapPaths {
// Every captured path must be a STRICT descendant of oldNs. Requiring the trailing "/" also
// catches p == oldNs (the namespace root itself — F-3a-3), which would otherwise map to a junk
// placement nesting the whole old namespace under the live root.
if !strings.HasPrefix(p, oldNs+"/") {
return nil, fmt.Errorf("a pillanatkép egy útvonala a névtéren kívülre mutat: %s", p)
}
rel := strings.TrimPrefix(p, oldNs+"/")
for _, seg := range strings.Split(rel, "/") {
if seg == ".." {
return nil, fmt.Errorf("a pillanatkép egy útvonala érvénytelen (..): %s", p)
}
}
isUnit := rel == "backups/primary/"+stack
if !isUnit && (rel == "backups" || strings.HasPrefix(rel, "backups/")) {
return nil, fmt.Errorf("nem-egység útvonal a fenntartott backups zónában: %s", p)
}
out = append(out, placement{
src: filepath.Join(scratch, p), // SP-3.1: abs source reconstructed under the target
dst: filepath.Join(liveNsRoot, rel),
isUnit: isUnit,
})
}
return out, nil
}
// placeCopier returns the place-to-live missing-only merge (nil seam → rsyncRestoreMissing, the
// `-a --ignore-existing` additive copy). NEVER rsyncMirror (--delete).
func (m *Manager) placeCopier() func(src, dst string) (int, error) {
if m.offboxPlaceCopier != nil {
return m.offboxPlaceCopier
}
return rsyncRestoreMissing
}
// PlaceOffsiteRestore places a COMPLETED full-scratch restore into the app's live locations via a
// missing-only merge (§7.3), so the SQ3 immich case is restorable from offsite alone. The recovery
// unit is placed ONLY if the live unit is ABSENT (never overwrites a local unit); every other path is
// merged missing-only. Does NOT deploy/start anything — RecreateStackFromUnit / the restore flow owns
// that. Single-flight. Requires a completed full scratch (deterministic path + existence check).
func (m *Manager) PlaceOffsiteRestore(ctx context.Context, stack string) error {
if !m.OffboxConfigured() {
return fmt.Errorf("off-box backup not configured")
}
if !isSafeStackName(stack) {
return fmt.Errorf("invalid stack name")
}
if err := m.acquireRunning(); err != nil {
return fmt.Errorf("egy másik mentési/visszaállítási művelet már fut")
}
defer m.releaseRunning()
scratch, _, err := m.offboxRestoreScratchDir(stack)
if err != nil {
return err
}
if _, sErr := os.Stat(scratch); sErr != nil {
return fmt.Errorf("nincs előkészített teljes visszaállítás — futtass előbb egy teljes visszaállítást")
}
id, paths, err := m.offboxLatestSnapshot(ctx, stack)
if err != nil {
return err
}
_ = id
// F-3a-1a: the live target uses the RAW HDD path (mirrors offboxCaptureSet). NOT AppNamespaceRoot —
// its systemDataPath fallback would merge userdata onto the SSD system namespace. Empty HDD ⇒
// undeployed ⇒ refuse: the app must be restored first, then its data placed under its live drive.
hdd := ""
if m.stackProvider != nil {
hdd = strings.TrimSpace(m.stackProvider.GetStackHDDPath(stack))
}
if hdd == "" {
return fmt.Errorf("a(z) %s nincs telepítve — előbb állítsd helyre az alkalmazást, utána az adatokat", stack)
}
liveNs := m.namespaceRoot(hdd)
// F-3a-1b: headroom gate — a missing-only merge copies at most the scratch size; refuse before any
// copy if the live drive lacks that (conservative — scratch and live often share a drive).
if free, need := m.offboxFree()(liveNs), m.offboxSize()(scratch); free < need {
return fmt.Errorf("Nincs elég szabad hely a visszaállításhoz (%s szükséges, %s szabad).", humanizeBytes(need), humanizeBytes(free))
}
placements, err := mapOffsiteRestorePaths(paths, stack, scratch, liveNs)
if err != nil {
return err // whole-placement refusal (no partial writes)
}
// F-3a-4: stat pre-pass over EVERY placement BEFORE the first copy — an incomplete scratch (e.g. a
// unit-only restore, userdata srcs absent) refuses with ZERO copies, making "no partial writes" true.
for _, pl := range placements {
if _, sErr := os.Stat(pl.src); sErr != nil {
return fmt.Errorf("a teljes visszaállítás hiányos (%s nincs meg) — futtass előbb egy teljes visszaállítást", filepath.Base(pl.src))
}
}
copier := m.placeCopier()
var placed int
for _, pl := range placements {
if pl.isUnit {
if _, liveErr := os.Stat(pl.dst); liveErr == nil {
m.logger.Printf("[INFO] [offbox] place %s: live recovery unit present — not overwriting", stack)
continue // never overwrite a local unit
}
}
n, cErr := copier(pl.src, pl.dst)
if cErr != nil {
return fmt.Errorf("a(z) %s helyreállítása sikertelen: %w", stack, cErr) // scratch KEPT for retry
}
placed += n
}
// F-3a-2: on FULL success, remove the scratch best-effort (OffboxFullScratchReady then turns false →
// the place button disappears). A failed placement returned above, keeping the scratch for a retry.
if rmErr := os.RemoveAll(scratch); rmErr != nil {
m.logger.Printf("[WARN] [offbox] place %s: scratch cleanup failed (harmless): %v", stack, rmErr)
} else {
m.logger.Printf("[INFO] [offbox] place %s: scratch removed after successful placement", stack)
}
m.logger.Printf("[INFO] [offbox] placed %s from offsite scratch: %d file(s) merged (missing-only)", stack, placed)
return nil
}