DataOps-platform/deployment/compose/backup.sh

#!/usr/bin/env bash

set -euo pipefail

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
# shellcheck source=wp13-common.sh
source "${SCRIPT_DIR}/wp13-common.sh"

usage() {
  cat <<'EOF'
Usage: backup.sh --output-dir PATH --confirm-maintenance-window [options]

Create a logical PostgreSQL export and a consistent cold snapshot of every
Compose-managed volume. The project is restarted even if archiving fails.

Options:
  --output-dir PATH              New backup directory.
  --compose-file PATH            Compose file.
  --env-file PATH                Optional Compose environment file.
  --project-name NAME            Source project (default: dataops-test).
  --confirm-maintenance-window   Required acknowledgement of bounded downtime.
  --allow-degraded               Permit backup when a container is not healthy.
  --timeout SECONDS              Restart health timeout (default: 600).
  --help                         Show this help.
EOF
}

ROOT="$(wp13_default_root)"
OUTPUT_DIR=""
COMPOSE_FILE=""
ENV_FILE=""
PROJECT_NAME="dataops-test"
CONFIRM_MAINTENANCE=0
ALLOW_DEGRADED=0
TIMEOUT_SECONDS=600

while [[ $# -gt 0 ]]; do
  case "$1" in
    --output-dir) OUTPUT_DIR="$2"; shift 2 ;;
    --compose-file) COMPOSE_FILE="$2"; shift 2 ;;
    --env-file) ENV_FILE="$2"; shift 2 ;;
    --project-name) PROJECT_NAME="$2"; shift 2 ;;
    --confirm-maintenance-window) CONFIRM_MAINTENANCE=1; shift ;;
    --allow-degraded) ALLOW_DEGRADED=1; shift ;;
    --timeout) TIMEOUT_SECONDS="$2"; shift 2 ;;
    --help|-h) usage; exit 0 ;;
    *) wp13_die "unknown argument: $1" ;;
  esac
done

[[ -n "${OUTPUT_DIR}" ]] || wp13_die "--output-dir is required"
[[ "${CONFIRM_MAINTENANCE}" -eq 1 ]] \
  || wp13_die "--confirm-maintenance-window is required"
COMPOSE_FILE="${COMPOSE_FILE:-${ROOT}/deploy/docker/docker-compose.yml}"
[[ -f "${COMPOSE_FILE}" ]] || wp13_die "compose file not found: ${COMPOSE_FILE}"
if [[ -e "${OUTPUT_DIR}" && -n "$(find "${OUTPUT_DIR}" -mindepth 1 -maxdepth 1 -print -quit 2>/dev/null)" ]]; then
  wp13_die "backup destination must be empty: ${OUTPUT_DIR}"
fi
mkdir -p "${OUTPUT_DIR}/volumes"
OUTPUT_DIR="$(cd "${OUTPUT_DIR}" && pwd)"

wp13_require_command docker
[[ -n "$(wp13_project_running_ids "${PROJECT_NAME}")" ]] \
  || wp13_die "source project is not running: ${PROJECT_NAME}"
if [[ "${ALLOW_DEGRADED}" -eq 0 ]]; then
  wp13_wait_for_project "${PROJECT_NAME}" 5
fi

STARTED_AT="$(date -u '+%Y-%m-%dT%H:%M:%SZ')"
wp13_log "creating logical PostgreSQL export"
wp13_compose "${COMPOSE_FILE}" "${PROJECT_NAME}" "${ENV_FILE}" \
  exec -T postgres pg_dumpall -U dataops \
  >"${OUTPUT_DIR}/postgres-all.sql"

wp13_compose "${COMPOSE_FILE}" "${PROJECT_NAME}" "${ENV_FILE}" \
  exec -T backend alembic -c alembic.ini current \
  >"${OUTPUT_DIR}/alembic-current.txt"

wp13_compose "${COMPOSE_FILE}" "${PROJECT_NAME}" "${ENV_FILE}" \
  config --images | LC_ALL=C sort -u \
  >"${OUTPUT_DIR}/images.txt"

VOLUME_NAMES="$(
  docker volume ls \
    --filter "label=com.docker.compose.project=${PROJECT_NAME}" \
    --format '{{.Name}}' \
    | LC_ALL=C sort
)"
[[ -n "${VOLUME_NAMES}" ]] || wp13_die "no Compose-managed volumes found"

PROJECT_STOPPED=0
restart_source() {
  if [[ "${PROJECT_STOPPED}" -eq 1 ]]; then
    wp13_log "restarting source project ${PROJECT_NAME}"
    if wp13_compose "${COMPOSE_FILE}" "${PROJECT_NAME}" "${ENV_FILE}" start >/dev/null; then
      wp13_wait_for_project "${PROJECT_NAME}" "${TIMEOUT_SECONDS}" || true
    fi
  fi
}
trap restart_source EXIT

wp13_log "stopping source project for consistent volume snapshots"
wp13_compose "${COMPOSE_FILE}" "${PROJECT_NAME}" "${ENV_FILE}" stop >/dev/null
PROJECT_STOPPED=1

{
  printf 'format=wp13-backup-v1\n'
  printf 'source_project=%s\n' "${PROJECT_NAME}"
  printf 'started_at_utc=%s\n' "${STARTED_AT}"
  printf 'compose_sha256=%s\n' "$(wp13_sha256_file "${COMPOSE_FILE}")"
  printf 'alembic_current=%s\n' "$(awk 'NF {print $1; exit}' "${OUTPUT_DIR}/alembic-current.txt")"
} >"${OUTPUT_DIR}/backup-manifest.txt"

while IFS= read -r volume_name; do
  [[ -n "${volume_name}" ]] || continue
  logical_name="$(
    docker volume inspect \
      -f '{{index .Labels "com.docker.compose.volume"}}' \
      "${volume_name}"
  )"
  [[ -n "${logical_name}" && "${logical_name}" != "<no value>" ]] \
    || wp13_die "volume is missing Compose logical label: ${volume_name}"
  archive_name="$(wp13_safe_archive_name "${logical_name}").tar.gz"
  wp13_log "archiving volume ${logical_name}"
  docker run --rm \
    -v "${volume_name}:/volume:ro" \
    -v "${OUTPUT_DIR}/volumes:/backup" \
    postgres:16-alpine \
    sh -c "cd /volume && tar -czf '/backup/${archive_name}' ."
  printf 'volume=%s|archive=volumes/%s|source_name=%s\n' \
    "${logical_name}" "${archive_name}" "${volume_name}" \
    >>"${OUTPUT_DIR}/backup-manifest.txt"
done <<<"${VOLUME_NAMES}"

wp13_compose "${COMPOSE_FILE}" "${PROJECT_NAME}" "${ENV_FILE}" start >/dev/null
PROJECT_STOPPED=0
wp13_wait_for_project "${PROJECT_NAME}" "${TIMEOUT_SECONDS}"

printf 'completed_at_utc=%s\n' "$(date -u '+%Y-%m-%dT%H:%M:%SZ')" \
  >>"${OUTPUT_DIR}/backup-manifest.txt"
wp13_write_checksums "${OUTPUT_DIR}"
wp13_log "backup completed: ${OUTPUT_DIR}"