web_hghl_h5/deploy_prod.sh

#!/bin/bash
# 黄果黑料正式服：构建 → Linux 运行包 + MD5 → 上传 → 切换 next/ → Supervisor 重启 → 健康检查（失败回滚）。
# 本地打包沿用上一版（对齐 hkdj-web）；服务器沿用现有结构：next/ + next.prev/，Supervisor 程序 web_hghl_h5（端口以其配置为准）。
set -euo pipefail

SCRIPT_DIR=$(cd "$(dirname "$0")" && pwd)
cd "$SCRIPT_DIR"
HOST="${DEPLOY_HOST:-root@172.104.166.87}"
SSH_PORT="${DEPLOY_SSH_PORT:-22}"
REMOTE_DIR="${DEPLOY_DIR:-/server/web_hghl/h5}"
PROGRAM="${DEPLOY_PROGRAM:-web_hghl_h5}"
TARGET_ARCH="${DEPLOY_TARGET_ARCH:-x64}"
TARGET_LIBC="${DEPLOY_TARGET_LIBC:-glibc}"
PACKAGE_NAME=web_hghl_h5.tar.gz
MD5_NAME=web_hghl_h5.md5
PACKAGE="$SCRIPT_DIR/$PACKAGE_NAME"
MD5_FILE="$SCRIPT_DIR/$MD5_NAME"
RELEASE="$SCRIPT_DIR/dist/production-linux-$TARGET_ARCH-$TARGET_LIBC"
PACKAGE_ONLY=0
DRY_RUN=0

fail() { printf '错误：%s\n' "$*" >&2; exit 1; }
step() { printf '\n▸ %s\n' "$*"; }

for argument in "$@"; do
  case "$argument" in
    --) ;;
    --dry-run) DRY_RUN=1 ;;
    --package-only) PACKAGE_ONLY=1 ;;
    -h|--help)
      printf '用法：%s [--dry-run | --package-only]\n' "$0"
      printf '默认上传 %s:%s/，切换 next/ 后重启 Supervisor 程序 %s。\n' "$HOST" "$REMOTE_DIR" "$PROGRAM"
      printf '可覆盖 DEPLOY_HOST、DEPLOY_SSH_PORT、DEPLOY_DIR、DEPLOY_PROGRAM、DEPLOY_TARGET_ARCH、DEPLOY_TARGET_LIBC。\n'
      exit 0 ;;
    *) fail "不支持的参数 ${argument}；使用 --help 查看用法" ;;
  esac
done

[[ "$HOST" =~ ^[a-zA-Z0-9_][a-zA-Z0-9_.@-]*$ ]] || fail 'DEPLOY_HOST 格式无效'
[[ "$REMOTE_DIR" =~ ^/[a-zA-Z0-9_-]+(/[a-zA-Z0-9_-]+)*/?$ ]] || fail 'DEPLOY_DIR 必须是普通绝对路径'
REMOTE_DIR="${REMOTE_DIR%/}"
[[ "$PROGRAM" =~ ^[a-zA-Z0-9_-]+$ ]] || fail 'DEPLOY_PROGRAM 格式无效'
[[ "$SSH_PORT" =~ ^[0-9]+$ ]] && [ "$SSH_PORT" -ge 1 ] && [ "$SSH_PORT" -le 65535 ] || fail 'SSH 端口必须在 1–65535 内'
case "$TARGET_ARCH" in x64|arm64) ;; *) fail 'DEPLOY_TARGET_ARCH 仅支持 x64 或 arm64' ;; esac
case "$TARGET_LIBC" in glibc|musl) ;; *) fail 'DEPLOY_TARGET_LIBC 仅支持 glibc 或 musl' ;; esac

for tool in node pnpm tar; do
  command -v "$tool" >/dev/null || fail "缺少命令 $tool"
done
node -e 'const [a,b]=process.versions.node.split(".").map(Number); if(a<20 || a===20 && b<9) process.exit(1)' || fail '需要 Node.js >=20.9，推荐 22'
command -v md5sum >/dev/null || command -v md5 >/dev/null || fail '缺少 md5sum 或 macOS md5'
for file in config/production.env package.json scripts/prepare-production-runtime.ts; do
  [ -s "$file" ] || fail "缺少 $file"
done
[ -x node_modules/.bin/tsx ] || fail '请先执行 pnpm install'

if [ "$DRY_RUN" = 1 ]; then
  printf 'DRY RUN：正式环境；Linux %s/%s；只核对，不构建、不连接服务器。\n' "$TARGET_ARCH" "$TARGET_LIBC"
  node --input-type=module -e 'import {readFileSync} from "node:fs"; import {parse} from "dotenv"; const c=parse(readFileSync("config/production.env")); console.log(`  PORT=${c.PORT} API=${c.API_PROXY_TARGET||c.NEXT_PUBLIC_API_BASE_URL} SITE=${c.NEXT_PUBLIC_SITE_URL}`)'
  printf '  pnpm build → bundle → bundle:linux -- %s %s → %s + %s\n' "$TARGET_ARCH" "$TARGET_LIBC" "$PACKAGE_NAME" "$MD5_NAME"
  if [ "$PACKAGE_ONLY" = 0 ]; then
    printf '  scp → %s:%s/；解压到 next.new，保留服务器上的 *.local.env\n' "$HOST" "$REMOTE_DIR"
    printf '  next → next.prev，next.new → next；不改动已有 Supervisor 配置，restart %s；/healthz 失败则回滚\n' "$PROGRAM"
  fi
  exit 0
fi

if [ "$PACKAGE_ONLY" = 0 ]; then
  for tool in ssh scp; do command -v "$tool" >/dev/null || fail "缺少命令 $tool"; done
fi
SSH=(ssh -p "$SSH_PORT" -o ConnectTimeout=15 "$HOST")

# 同仓库只允许一次构建/上传，避免并发覆盖同名包。
mkdir -p "$SCRIPT_DIR/dist"
DEPLOY_LOCK="$SCRIPT_DIR/dist/.production-deploy.lock"
mkdir "$DEPLOY_LOCK" 2>/dev/null || fail "已有正式发布进程或残留锁：${DEPLOY_LOCK}；请先确认该进程已结束"
trap 'rmdir "$DEPLOY_LOCK" 2>/dev/null || true' EXIT
trap 'exit 130' INT
trap 'exit 143' TERM

step '构建并整理正式环境 standalone'
pnpm run build
pnpm run bundle
step "准备 Linux $TARGET_ARCH/$TARGET_LIBC 运行包"
pnpm run bundle:linux -- "$TARGET_ARCH" "$TARGET_LIBC"

for file in start.mjs server.js .next/BUILD_ID config/production.env runtime-target.json; do
  [ -s "$RELEASE/$file" ] || fail "正式运行包缺少 $file"
done
for directory in node_modules .next/static public; do
  [ -d "$RELEASE/$directory" ] || fail "正式运行包缺少目录 $directory"
done
# 软链必须是相对路径，指向本机绝对路径的软链上传到服务器后会失效（Cannot find module）
ABSOLUTE_LINKS=$(find "$RELEASE" -type l -lname '/*' | head -n 5)
[ -z "$ABSOLUTE_LINKS" ] || fail "运行包含有指向本机绝对路径的软链：
$ABSOLUTE_LINKS"
BROKEN_LINKS=$(find -L "$RELEASE" -type l | head -n 5)
[ -z "$BROKEN_LINKS" ] || fail "运行包含有失效软链：
$BROKEN_LINKS"

step "打包 $PACKAGE_NAME"
# -C 保持产物直接位于 tar 根部；禁止 AppleDouble、环境覆盖和日志混入。
COPYFILE_DISABLE=1 COPY_EXTENDED_ATTRIBUTES_DISABLE=1 tar \
  --no-xattrs \
  --exclude='._*' --exclude='.DS_Store' --exclude='.env*' \
  --exclude='*.local.env' --exclude='*.log' \
  -czf "$PACKAGE" -C "$RELEASE" .
chmod 600 "$PACKAGE"
if command -v md5sum >/dev/null; then
  md5sum "$PACKAGE" | awk '{print $1}' > "$MD5_FILE"
else
  md5 -q "$PACKAGE" > "$MD5_FILE"
fi
tar -tzf "$PACKAGE" >/dev/null
node -e 'console.log(`运行包已生成，体积 ${(require("node:fs").statSync(process.argv[1]).size / 1024 / 1024).toFixed(2)} MiB`)' "$PACKAGE"
if [ "$PACKAGE_ONLY" = 1 ]; then exit 0; fi

step "上传到 $HOST:$REMOTE_DIR/"
"${SSH[@]}" "mkdir -p '$REMOTE_DIR'"
scp -P "$SSH_PORT" "$PACKAGE" "$MD5_FILE" "$HOST:$REMOTE_DIR/"

step "切换版本并重启 $PROGRAM"
"${SSH[@]}" "bash -s -- '$REMOTE_DIR' '$PROGRAM' '$PACKAGE_NAME' '$MD5_NAME'" <<'REMOTE'
set -euo pipefail
DIR=$1 PROGRAM=$2 PACKAGE=$3 MD5=$4
cd "$DIR"
for tool in md5sum tar curl supervisorctl; do
  command -v "$tool" >/dev/null || { echo "服务器缺少命令 $tool"; exit 1; }
done
# ssh 非登录 shell 不加载 nvm 等环境：依次找 PATH、同机花魁的 Supervisor 配置、nvm、常见安装目录
find_node() {
  command -v node 2>/dev/null && return 0
  local from_conf
  from_conf=$(sed -n 's/^command=\([^ ]*node\) .*/\1/p' /etc/supervisor/conf.d/*.conf 2>/dev/null | head -n 1)
  [ -n "$from_conf" ] && [ -x "$from_conf" ] && { echo "$from_conf"; return 0; }
  local candidate
  for candidate in /root/.nvm/versions/node/*/bin/node /usr/local/bin/node /usr/bin/node /opt/node*/bin/node; do
    [ -x "$candidate" ] && { echo "$candidate"; return 0; }
  done
  return 1
}
NODE_BIN=$(find_node) || { echo "服务器上找不到 node（Node.js >=20.9），请先安装"; exit 1; }
echo "使用 node：$NODE_BIN（$("$NODE_BIN" -v)）"
[ "$(md5sum "$PACKAGE" | awk '{print $1}')" = "$(tr -d '[:space:]' < "$MD5")" ] || { echo "MD5 校验失败"; exit 1; }

# 目录沿用服务器现有结构（与 deploy/up_next_h5.sh 一致）：next/ 为当前版本，next.prev/ 为上一版
CURRENT=next PREVIOUS=next.prev STAGING=next.new
CONF="/etc/supervisor/conf.d/$PROGRAM.conf"

# 自愈：之前误写过指向 app/ 的配置时，恢复最近一次备份（原配置）
if [ -f "$CONF" ] && grep -qx "directory=$DIR/app" "$CONF"; then
  LATEST_BACKUP=$(ls -t "$CONF".bak.* 2>/dev/null | head -n 1 || true)
  if [ -n "$LATEST_BACKUP" ] && grep -qx "directory=$DIR/$CURRENT" "$LATEST_BACKUP"; then
    cp -p "$LATEST_BACKUP" "$CONF"
    echo "已从 $LATEST_BACKUP 恢复原 Supervisor 配置"
  fi
fi
# 清理之前误建的 app 系列目录（服务器原本没有）
rm -rf app app.new app.prev app.failed

rm -rf "$STAGING" && mkdir "$STAGING"
tar -xzf "$PACKAGE" -C "$STAGING"
[ -s "$STAGING/start.mjs" ] && [ -s "$STAGING/.next/BUILD_ID" ] || { echo "运行包内容不完整"; exit 1; }
# 服务器上的本地覆盖配置优先于包内文件，且不随发布丢失。
if [ -d "$CURRENT" ]; then
  for file in "$CURRENT"/config/*.local.env; do [ -f "$file" ] && cp -p "$file" "$STAGING/config/"; done
fi

# 已有 Supervisor 配置不改动；端口以配置里的 PORT 为准（进程环境变量优先于包内 production.env）
PACKAGE_PORT=$(sed -n 's/^PORT=//p' "$STAGING/config/production.env" | tail -n 1)
if [ -f "$CONF" ]; then
  PORT=$(grep -o 'PORT="[0-9]*"' "$CONF" | head -n 1 | grep -o '[0-9]*' || true)
  PORT=${PORT:-$PACKAGE_PORT}
  [ "$PORT" = "$PACKAGE_PORT" ] || echo "提示：Supervisor 配置端口 $PORT 与包内 PORT=$PACKAGE_PORT 不同，以配置为准"
else
  PORT=${PACKAGE_PORT:-20123}
  cat > "$CONF" <<CONF
[program:$PROGRAM]
directory=$DIR/$CURRENT
command=$NODE_BIN $DIR/$CURRENT/start.mjs
environment=HOSTNAME="0.0.0.0",PORT="$PORT",APP_ENV="production",NODE_ENV="production",PATH="/usr/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin"
autostart=true
autorestart=true
startsecs=5
stopasgroup=true
killasgroup=true
stopwaitsecs=15
stdout_logfile=/var/log/$PROGRAM.log
stderr_logfile=/var/log/${PROGRAM}_error.log
stdout_logfile_maxbytes=500MB
stdout_logfile_backups=5
CONF
  echo "首次部署：已创建 Supervisor 配置 $CONF"
fi

rm -rf "$PREVIOUS"
[ -d "$CURRENT" ] && mv "$CURRENT" "$PREVIOUS"
mv "$STAGING" "$CURRENT"
supervisorctl reread >/dev/null && supervisorctl update >/dev/null
supervisorctl restart "$PROGRAM" >/dev/null || true
healthy() {
  for _ in $(seq 1 30); do
    if curl -fsS -m 4 "http://127.0.0.1:$PORT/healthz" 2>/dev/null | grep -q '"status":"ok"'; then return 0; fi
    sleep 2
  done
  return 1
}
diagnose() {
  echo "---- 诊断信息 ----"
  supervisorctl status "$PROGRAM" || true
  local pid
  pid=$(supervisorctl pid "$PROGRAM" 2>/dev/null | grep -E '^[0-9]+$' || true)
  if [ -n "$pid" ] && command -v ss >/dev/null; then
    echo "进程 pid=$pid 实际监听："
    ss -ltnpH 2>/dev/null | grep "pid=$pid," | awk '{print "  " $4}' || true
  fi
  echo "最近错误日志："
  tail -n 30 "/var/log/${PROGRAM}_error.log" 2>/dev/null || true
  echo "最近输出日志："
  tail -n 200 "/var/log/$PROGRAM.log" 2>/dev/null | grep -vE "^\s|^\[home/base\]|^}" | tail -n 20 || true
  echo "------------------"
}
if healthy && curl -fsS -m 10 -o /dev/null "http://127.0.0.1:$PORT/"; then
  supervisorctl status "$PROGRAM"
  echo "DEPLOY_OK BUILD_ID=$(cat "$CURRENT/.next/BUILD_ID") PORT=$PORT"
  exit 0
fi
echo "新版本健康检查失败（127.0.0.1:$PORT/healthz 无响应）"
diagnose
if [ -d "$PREVIOUS" ]; then
  echo "回滚到 $PREVIOUS"
  rm -rf next.failed && mv "$CURRENT" next.failed && mv "$PREVIOUS" "$CURRENT"
  supervisorctl restart "$PROGRAM" >/dev/null || true
  healthy && echo "已回滚并恢复服务；失败版本保留在 next.failed" || echo "回滚后服务仍未恢复，请立即人工检查"
else
  echo "没有可回滚的上一版，新版本保留在 $CURRENT/"
fi
exit 1
REMOTE
printf '\n正式服发布完成。\n'