homelab-infra/ops/restore-tests/immich-restore-test.sh

#!/bin/bash
set -euo pipefail

# Immich Restore Smoke Test
#
# Nicht-destruktiver Restore-Smoke-Test fuer Immich.
# - liest immich.dump aus dem produktiven Borg-Archiv
# - importiert in eine isolierte Test-Postgres-Instanz mit gleichem Image
#   wie Produktion (tensorchord/pgvecto-rs)
# - startet einen isolierten Immich-Server-Container ohne Traefik und
#   ohne ML-Container
# - prueft Login-Page und Asset-Anzahl aus DB
# - bereinigt anschliessend
#
# Produktiver Immich-Stack wird NICHT angefasst.
# Produktive Foto-Pfade unter /mnt/user/photos/* werden NICHT gemountet.

SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
. "$SCRIPT_DIR/common.sh"

WHATIF=0
KEEP_DATA=0
for arg in "$@"; do
  case "$arg" in
    --what-if) WHATIF=1 ;;
    --keep-data) KEEP_DATA=1 ;;
    *) echo "Unknown argument: $arg" >&2; exit 1 ;;
  esac
done

RESTORE_ROOT="/mnt/user/backups/restore-lab/immich"
REPORT_ROOT="/mnt/user/backups/restore-reports"
EXTRACT_DIR="$BORG_RESTORE_HOST_ROOT/immich-extract"
COMPOSE_FILE="$SCRIPT_DIR/immich-compose.test.yml"
REPORT_FILE="$REPORT_ROOT/immich-$(date +%F).md"

if [ "$WHATIF" -eq 1 ]; then
  cat <<EOF
Immich restore test
Mode: WhatIf
RestoreRoot: $RESTORE_ROOT
ReportRoot: $REPORT_ROOT
Expected Borg source paths:
- local/borg-dumps/latest/immich.dump
Planned isolation:
- Test-Postgres: tensorchord/pgvecto-rs:pg14-v0.2.0
- Test-Redis: redis:7.4-alpine (rebuildbar, kein Restore)
- Test-Server: ghcr.io/immich-app/immich-server:release (Image-Pin wie Produktion)
- ML-Container bewusst weggelassen
- Test-Upload: leer, unter $RESTORE_ROOT/upload
- Productive photo paths NOT mounted: /mnt/user/photos/immich, /mnt/user/photos/family_archive
- Test endpoint: 127.0.0.1:12283 (no Traefik, no public domain)
Smoke-Test:
- Test-Postgres healthy
- pg_restore -Fc -> immich.dump
- HTTP 200/302/3xx von 127.0.0.1:12283
- Asset-Count aus DB
EOF
  exit 0
fi

require_cmd docker
require_cmd curl
require_path "$BORG_PASSPHRASE_FILE_DEFAULT"
require_path "$COMPOSE_FILE"

cleanup() {
  cleanup_compose "$COMPOSE_FILE"
  if [ "$KEEP_DATA" -ne 1 ]; then
    rm -rf "$RESTORE_ROOT"
  fi
  rm -rf "$EXTRACT_DIR"
}
trap cleanup EXIT

rm -rf "$EXTRACT_DIR" "$RESTORE_ROOT"
mkdir -p "$RESTORE_ROOT/postgres" "$RESTORE_ROOT/upload" "$RESTORE_ROOT/dumps/latest"

archive="$(latest_archive_name)"
repo="$(borg_repo_url)"

borg_extract "/restore/immich-extract" \
  "local/borg-dumps/latest/immich.dump"

mv "$EXTRACT_DIR/local/borg-dumps/latest/immich.dump" "$RESTORE_ROOT/dumps/latest/immich.dump"

# Stufe 1: Test-Postgres und Test-Redis starten
docker compose -f "$COMPOSE_FILE" up -d \
  restoretest-immich-postgres restoretest-immich-redis >/dev/null

# Warten auf Postgres ready
until docker exec restoretest-immich-postgres pg_isready -U immich -d immich >/dev/null 2>&1; do
  sleep 2
done

# Stufe 2: Dump in Test-Postgres importieren
# Hinweis: pg_restore mit --clean --if-exists, damit die Operation idempotent ist.
# --no-owner / --no-privileges, weil im Test-Postgres kein produktiver User existiert.
docker exec -i restoretest-immich-postgres \
  pg_restore -U immich -d immich --clean --if-exists --no-owner --no-privileges \
  < "$RESTORE_ROOT/dumps/latest/immich.dump"

# Stufe 3: Immich-Server starten (ohne ML)
docker compose -f "$COMPOSE_FILE" up -d restoretest-immich-server >/dev/null

# Immich-Server braucht beim ersten Start einige Sekunden fuer DB-Migrations-Checks.
# Wir geben ihm bis zu 120s und pollen den HTTP-Endpunkt.
http_status=""
for _ in $(seq 1 60); do
  http_status="$(curl -s -o /tmp/immich-body.html -w '%{http_code}' -L http://127.0.0.1:12283 || true)"
  if [ "$http_status" = "200" ] || [ "$http_status" = "302" ] || [ "$http_status" = "303" ]; then
    break
  fi
  sleep 2
done

# Body-Check: Immich-UI hat typische Marker. Wir matchen tolerant.
body_check="ok"
if ! grep -qiE "immich|login|signin" /tmp/immich-body.html 2>/dev/null; then
  body_check="missing-marker"
fi

# Asset-Count aus DB. Wenn die Spalte nicht existiert (Schema-Drift),
# wird das im Report sichtbar gemacht statt das Skript zu killen.
asset_count="$(docker exec restoretest-immich-postgres \
  psql -U immich -d immich -tAc "select count(*) from assets;" 2>/dev/null \
  | tr -d '[:space:]' || true)"
if [ -z "$asset_count" ]; then
  asset_count="n/a"
fi

# User-Count als zusaetzlicher DB-Sanity-Check
user_count="$(docker exec restoretest-immich-postgres \
  psql -U immich -d immich -tAc "select count(*) from users;" 2>/dev/null \
  | tr -d '[:space:]' || true)"
if [ -z "$user_count" ]; then
  user_count="n/a"
fi

write_report "$REPORT_FILE" <<EOF
# Immich Restore Test Report - $(date +%F)

- Service: \`immich\`
- Source repo: \`$repo\`
- Archive: \`$archive\`
- Restore root: \`$RESTORE_ROOT\`
- Test containers:
  - \`restoretest-immich-server\`
  - \`restoretest-immich-postgres\` (tensorchord/pgvecto-rs:pg14-v0.2.0)
  - \`restoretest-immich-redis\`
- Test endpoint: \`http://127.0.0.1:12283\`
- ML container: deliberately omitted
- Result: \`SUCCESS\`

## Checks

- Borg extract of \`immich.dump\`: \`ok\`
- Dump import into isolated Postgres: \`ok\`
- HTTP status after redirect: \`$http_status\`
- Login page marker: \`$body_check\`
- Asset count in test DB: \`$asset_count\`
- User count in test DB: \`$user_count\`

## Notes

- Test ran without Traefik and without the productive domain.
- Productive photo paths under /mnt/user/photos/* were NOT mounted.
- Test data was cleaned after success: \`$([ "$KEEP_DATA" -eq 1 ] && echo no || echo yes)\`
- Restore-Quelle Dump: \`local/borg-dumps/latest/immich.dump\` aus aktuellem Borg-Archiv.
EOF

echo "Immich restore test ok -> $REPORT_FILE"