Momento/scripts/deploy-prod.sh

#!/usr/bin/env bash
# Déploiement prod sur 192.168.1.190 — exécuter SUR LE SERVEUR (runner docker-host).
set -euo pipefail

# Telegram notification function
telegram_notify() {
  local status="$1"
  local message="$2"
  local bot_token="${TELEGRAM_BOT_TOKEN:-}"
  local chat_id="${TELEGRAM_CHAT_ID:-}"

  if [ -z "$bot_token" ] || [ -z "$chat_id" ]; then
    echo "TELEGRAM_BOT_TOKEN or TELEGRAM_CHAT_ID not set — skipping notification"
    return 0
  fi

  local emoji="✅"
  if [ "$status" = "failure" ]; then
    emoji="❌"
  elif [ "$status" = "rollback" ]; then
    emoji="🔄"
  fi

  local commit_msg=$(git log -1 --pretty=format:"%s" 2>/dev/null || echo "unknown")
  # Escape markdown special characters in commit message
  commit_msg=$(echo "$commit_msg" | sed 's/[_*`[\]()~>#+=|{}.!\\-]/\\&/g')
  local commit_short="${GIT_COMMIT:0:8}"

  curl -s -X POST "https://api.telegram.org/bot${bot_token}/sendMessage" \
    -d "chat_id=${chat_id}" \
    -d "text=${emoji} Momento Deploy ${status}
Commit: ${commit_short}
Message: ${commit_msg}
Details: ${message}" \
    -d "parse_mode=Markdown" > /dev/null 2>&1 || echo "Telegram notification failed"
}

# Rollback functions
rollback_save_image() {
  echo "=== Saving current image for rollback ==="
  if docker images --format '{{.Repository}}:{{.Tag}}' | grep -q "^memento-memento-note:latest"; then
    docker tag memento-memento-note:latest memento-memento-note:rollback || echo "WARN: Failed to tag rollback image"
    echo "Rollback image saved: memento-memento-note:rollback"
  else
    echo "WARN: No current memento-note image found to save"
  fi
}

rollback_restore_image() {
  echo "=== Rolling back to previous image ==="
  if docker images --format '{{.Repository}}:{{.Tag}}' | grep -q "^memento-memento-note:rollback"; then
    docker tag memento-memento-note:rollback memento-memento-note:latest || echo "ERROR: Failed to restore rollback image"
    docker compose up -d --force-recreate memento-note || echo "ERROR: Failed to restart container"
    echo "Rollback completed"
  else
    echo "ERROR: No rollback image found"
    return 1
  fi
}

ROOT="${DEPLOY_ROOT:-/opt/memento}"
ARTIFACT_TGZ="${ARTIFACT_TGZ:-}"
EXPECTED_COMMIT="${EXPECTED_COMMIT:-}"

load_env_docker() {
  local env_file="$ROOT/.env.docker"
  [ -f "$env_file" ] || return 0
  # Parse ligne par ligne (ne crashe pas sur une valeur avec quote mal fermée,
  # contrairement à `source` qui exécute le fichier comme du bash).
  local line key val
  while IFS= read -r line || [ -n "$line" ]; do
    case "$line" in
      ''|\#*) continue ;;           # vide / commentaire
      *=*) ;;                        # doit contenir =
      *) continue ;;
    esac
    key="${line%%=*}"
    val="${line#*=}"
    key="${key%%[[:space:]]}"  # trim trailing whitespace du key
    # skip placeholders et valeurs vides
    [ -z "$val" ] && continue
    [ "$val" = "???" ] && continue
    # strip les guillemets entourants seulement s'ils sont équilibrés
    case "$val" in
      \"*\") val="${val#\"}"; val="${val%\"}" ;;
      \'*\') val="${val#\'}"; val="${val%\'}" ;;
    esac
    # valide le nom de variable (alnum + underscore, commence par une lettre)
    case "$key" in
      [A-Za-z_]*|[A-Za-z_]*) export "$key=$val" ;;
    esac
  done < "$env_file"
}

wait_for_postgres() {
  local pg_user="${POSTGRES_USER:-memento}"
  local pg_db="${POSTGRES_DB:-memento}"

  if docker compose up -d --wait postgres 2>/dev/null; then
    echo "Postgres healthy (compose --wait)"
    return 0
  fi

  docker compose up -d postgres

  for i in $(seq 1 30); do
    # Use credentials from inside the container (authoritative)
    if docker compose exec -T postgres sh -c 'pg_isready -U "$POSTGRES_USER" -d "$POSTGRES_DB"' >/dev/null 2>&1; then
      echo "Postgres ready (${i}/30)"
      return 0
    fi
    # Fallback: host env after load_env_docker
    if docker compose exec -T postgres pg_isready -U "$pg_user" -d "$pg_db" >/dev/null 2>&1; then
      echo "Postgres ready via host env (${i}/30)"
      return 0
    fi
    sleep 2
  done

  echo "Postgres not ready after 60s"
  docker compose ps postgres 2>/dev/null || true
  docker compose logs postgres --tail=50 2>/dev/null || true
  docker compose exec -T postgres sh -c 'pg_isready -U "$POSTGRES_USER" -d "$POSTGRES_DB"' 2>&1 || true
  return 1
}

# Health check configuration: 24 iterations × 5 seconds = 2 minutes total timeout
# This allows Next.js cold start time while keeping feedback fast
HEALTH_CHECK_MAX_ITERATIONS=24
HEALTH_CHECK_SLEEP_SECONDS=5

cd "$ROOT"

if [ -f "$ROOT/.env.docker" ]; then
  sed -i 's/\r$//' "$ROOT/.env.docker"
  # Docker Compose reads .env (not .env.docker) for ${...} interpolation.
  # Sync .env from .env.docker so POSTGRES_PASSWORD etc. are interpolated correctly.
  grep -v '???' "$ROOT/.env.docker" | grep -v '^#' > "$ROOT/.env"
fi

load_env_docker
git config --global --add safe.directory "$ROOT" 2>/dev/null || true
git fetch origin main
git reset --hard origin/main
export GIT_COMMIT="$(git rev-parse HEAD)"
echo "=== Deploy $GIT_COMMIT ==="

if [ -n "$EXPECTED_COMMIT" ] && [ "$GIT_COMMIT" != "$EXPECTED_COMMIT" ]; then
  echo "ERROR: git HEAD $GIT_COMMIT != expected $EXPECTED_COMMIT"
  exit 1
fi

wait_for_postgres || exit 1

docker compose exec -T postgres sh -c 'psql -U "$POSTGRES_USER" -d "$POSTGRES_DB" -c "CREATE EXTENSION IF NOT EXISTS vector;"' >/dev/null

if git diff --name-only HEAD~1 HEAD 2>/dev/null | grep -q '^memento-note/prisma/migrations/'; then
  DUMP_FILE="/opt/memento/backups/pre-migrate-$(date +%Y%m%d-%H%M%S).sql.gz"
  mkdir -p /opt/memento/backups
  docker compose exec -T postgres pg_dump -U "${POSTGRES_USER:-memento}" -d "${POSTGRES_DB:-memento}" | gzip > "$DUMP_FILE"
  DUMP_SIZE=$(stat -c%s "$DUMP_FILE")
  if [ "$DUMP_SIZE" -lt 1048576 ]; then
    echo "ERROR: Dump too small ($DUMP_SIZE bytes)"
    exit 1
  fi
  echo "Backup: $DUMP_FILE ($(( DUMP_SIZE / 1024 ))KB)"
else
  echo "No migration change — skip pre-migrate dump"
fi

export COMPOSE_DOCKER_CLI_BUILD=1 DOCKER_BUILDKIT=1
export GIT_COMMIT

# Save current image for rollback before building new one
rollback_save_image

if [ -n "$ARTIFACT_TGZ" ] && [ -f "$ARTIFACT_TGZ" ]; then
  echo "=== Fast image (CI artifact) ==="
  tar xzf "$ARTIFACT_TGZ" -C memento-note/
  export MEMENTO_DOCKERFILE=Dockerfile.prebuilt
  export MEMENTO_SOCKET_DOCKERFILE=Dockerfile.socket.prebuilt
else
  echo "ERROR: No CI artifact at $ARTIFACT_TGZ — full Docker build is not supported on this host."
  telegram_notify "failure" "Deploy aborted: missing prebuilt artifact"
  exit 1
fi

docker compose build memento-note
docker compose build memento-socket

if git diff --name-only HEAD~1 HEAD 2>/dev/null | grep -q '^mcp-server/'; then
  docker compose build mcp-server
fi

docker compose up -d --remove-orphans --force-recreate memento-note
docker compose up -d memento-socket
docker compose up -d mcp-server 2>/dev/null || true

# Monitoring stack updates are handled at the end of successful deployment

echo "=== Migrations (Prisma CLI via node, pas npx) ==="
if docker compose exec -T memento-note test -f ./node_modules/prisma/build/index.js 2>/dev/null; then
  docker compose exec -T memento-note node ./node_modules/prisma/build/index.js migrate deploy
else
  echo "WARN: prisma CLI absent de l'image — les migrations tournent au démarrage (entrypoint) ou via psql manuel"
fi

nginx -t 2>/dev/null && systemctl reload nginx 2>/dev/null || true

for i in $(seq 1 "$HEALTH_CHECK_MAX_ITERATIONS"); do
  BODY=$(docker compose exec -T memento-note node -e "require('http').get('http://localhost:3000/api/build-info',r=>{let d='';r.on('data',c=>d+=c);r.on('end',()=>{console.log(d);process.exit(0)})}).on('error',()=>process.exit(1))" 2>/dev/null || true)
  ACTUAL=$(echo "$BODY" | jq -r '.commit // empty' 2>/dev/null || true)
  if [ "$ACTUAL" = "$GIT_COMMIT" ]; then
    echo "OK build-info commit=$ACTUAL"
    echo "=== Updating monitoring stack ==="
    # Fix potential Docker directory/file mount corruption for metrics-token
    if [ -d monitoring/metrics-token ]; then
      echo "WARN: monitoring/metrics-token is a directory on host, removing it"
      rm -rf monitoring/metrics-token
      git checkout monitoring/metrics-token || echo "default-token-value" > monitoring/metrics-token
    fi

    load_env_docker
    if [ -n "${TELEGRAM_BOT_TOKEN:-}" ] && [ -n "${TELEGRAM_CHAT_ID:-}" ]; then
      echo "=== Starting Monitoring Stack (with Telegram bot) ==="
      docker compose -f monitoring/docker-compose.monitoring.yml --profile telegram up -d --remove-orphans 2>&1 || echo "WARN: Failed to bring up monitoring stack"
    else
      echo "=== Starting Monitoring Stack (without Telegram bot) ==="
      docker compose -f monitoring/docker-compose.monitoring.yml up -d --remove-orphans 2>&1 || echo "WARN: Failed to bring up monitoring stack"
    fi

    # Force recreation of Grafana, Prometheus and Alertmanager to apply config changes and fix mounts
    echo "=== Recreating Grafana, Prometheus and Alertmanager containers ==="
    docker compose -f monitoring/docker-compose.monitoring.yml up -d --force-recreate grafana prometheus alertmanager 2>&1 || true

    docker compose ps

    telegram_notify "success" "Deployment successful — app is healthy"
    exit 0
  fi
  sleep "$HEALTH_CHECK_SLEEP_SECONDS"
done

echo "ERROR: build-info=$ACTUAL expected=$GIT_COMMIT"
docker logs memento-web --tail=40

# Rollback on health check failure
rollback_restore_image
telegram_notify "rollback" "Health check failed — rolled back to previous version"

exit 1