web/scripts/lib-yh-compose-deploy.sh

# shellcheck shell=bash
# 由 pull-and-restart.sh / restart.sh 在定义好 ROOT、compose_cmd、run_sudo 之后 source。
# 统一策略：仅使用宿主机 Nginx；容器 yh_nginx 不再作为入口。

YH_COMPOSE_FILES="-f docker-compose.yml -f docker-compose.host-nginx.yml"

require_cmd() {
  local c="$1"
  command -v "$c" >/dev/null 2>&1 || {
    echo "错误: 缺少命令 $c，请先安装后重试。" >&2
    exit 1
  }
}

ensure_host_deploy_env() {
  require_cmd systemctl
  require_cmd ss
  require_cmd sed
  require_cmd awk
  require_cmd curl
}

host_nginx_online() {
  command -v systemctl >/dev/null 2>&1 || return 1
  systemctl is-active nginx >/dev/null 2>&1 && return 0
  systemctl is-active nginx.service >/dev/null 2>&1 && return 0
  return 1
}

force_release_port() {
  local p="$1"
  local victims
  victims="$(run_sudo ss -tlnp "sport = :$p" 2>/dev/null | sed -n 's/.*pid=\([0-9]\+\).*/\1/p' | awk '!seen[$0]++')"
  [ -z "$victims" ] && return 0

  for pid in $victims; do
    [ -z "$pid" ] && continue
    local comm
    comm="$(run_sudo sh -c "cat /proc/$pid/comm 2>/dev/null" || true)"
    if [ "$comm" = "nginx" ]; then
      continue
    fi
    echo "端口 $p 被非宿主机 nginx 进程占用（pid=$pid, comm=${comm:-unknown}），强制停止..."
    run_sudo kill -9 "$pid" 2>/dev/null || true
  done
}

force_release_host_ports() {
  # 先优先停掉所有发布了宿主机 80/443 的容器（最常见冲突源）
  if command -v docker >/dev/null 2>&1; then
    for cid in $(run_sudo docker ps --filter "publish=80" --format "{{.ID}}" 2>/dev/null); do
      [ -n "$cid" ] && run_sudo docker rm -f "$cid" 2>/dev/null || true
    done
    for cid in $(run_sudo docker ps --filter "publish=443" --format "{{.ID}}" 2>/dev/null); do
      [ -n "$cid" ] && run_sudo docker rm -f "$cid" 2>/dev/null || true
    done
  fi
  # 再兜底杀掉非 nginx 的占用进程
  force_release_port 80
  force_release_port 443
}

ensure_host_nginx_started() {
  ensure_host_deploy_env
  force_release_host_ports

  if host_nginx_online; then
    echo "宿主机 Nginx 在线，跳过启动。"
    return 0
  fi
  echo "宿主机 Nginx 未在线，尝试启动..."
  run_sudo systemctl start nginx 2>/dev/null || run_sudo systemctl start nginx.service
  run_sudo systemctl enable nginx 2>/dev/null || true
  if host_nginx_online; then
    echo "宿主机 Nginx 启动成功。"
    return 0
  fi
  echo "错误: 无法启动宿主机 Nginx，请检查 systemctl status nginx" >&2
  exit 1
}

# 停止本项目 Compose 栈（包含可能残留的 yh_nginx），不停止宿主机 Nginx。
yh_compose_down() {
  if [ ! -f "$ROOT/docker-compose.host-nginx.yml" ]; then
    compose_cmd down --remove-orphans 2>/dev/null || true
    return 0
  fi
  compose_cmd $YH_COMPOSE_FILES down --remove-orphans 2>/dev/null || true
}

# 仅启动业务容器，不再启动容器 yh_nginx。
yh_compose_up() {
  if [ ! -f "$ROOT/docker-compose.host-nginx.yml" ]; then
    echo "未找到 docker-compose.host-nginx.yml，使用默认 compose 启动业务容器。"
    compose_cmd up -d --force-recreate mongo api web admin
    return 0
  fi
  compose_cmd $YH_COMPOSE_FILES up -d --force-recreate mongo api web admin
}

# 从模板生成宿主机站点配置，并在配置检查通过后 reload（若离线则 start）。
yh_install_host_nginx_site_conf() {
  local domain="${NGINX_DOMAIN:-yuheng.yuxindazhineng.com}"
  local tpl="$ROOT/nginx/yuheng.host.conf"
  local out="/etc/nginx/conf.d/${domain}.conf"
  local ts
  ts="$(date +%Y%m%d%H%M%S)"

  if [ ! -f "$tpl" ]; then
    echo "未找到 $tpl，跳过宿主机站点配置生成。"
    return 0
  fi

  ensure_host_deploy_env

  # 同域名冲突兜底：将 conf.d 下其他包含相同 server_name 的配置下线，避免错误 upstream 继续生效导致 502。
  run_sudo sh -c "for f in /etc/nginx/conf.d/*.conf; do
    [ -f \"\$f\" ] || continue
    [ \"\$f\" = \"$out\" ] && continue
    if grep -Eq '^[[:space:]]*server_name[[:space:]]+[^;]*${domain}([[:space:];]|$)' \"\$f\"; then
      mv -f \"\$f\" \"\${f}.disabled_by_yh_${ts}\"
    fi
  done"

  mkdir -p "$ROOT/verify-root"
  sed "s|__VERIFY_ROOT__|$ROOT/verify-root|g" "$tpl" | run_sudo tee "$out" >/dev/null

  if ! run_sudo nginx -t 2>/dev/null; then
    echo "错误: 宿主机 nginx -t 失败，请检查 $out" >&2
    exit 1
  fi

  if host_nginx_online; then
    run_sudo systemctl reload nginx 2>/dev/null && echo "宿主机 Nginx 已重载（$out）。" || true
  else
    ensure_host_nginx_started
  fi
}

yh_post_deploy_healthcheck() {
  local domain="${NGINX_DOMAIN:-yuheng.yuxindazhineng.com}"
  local code=""
  local ok_web=0
  local ok_admin=0
  local ok_api=0

  # 先验证上游容器端口（带重试，避免容器刚起时瞬时连接拒绝）
  for _ in $(seq 1 20); do
    curl -fsS --max-time 3 http://127.0.0.1:9080/ >/dev/null && ok_web=1 && break
    sleep 1
  done
  [ "$ok_web" -eq 1 ] || {
    echo "错误: 前台上游 127.0.0.1:9080 不可用" >&2
    return 1
  }
  for _ in $(seq 1 20); do
    curl -fsS --max-time 3 http://127.0.0.1:9081/ >/dev/null && ok_admin=1 && break
    sleep 1
  done
  [ "$ok_admin" -eq 1 ] || {
    echo "错误: 后台上游 127.0.0.1:9081 不可用" >&2
    return 1
  }
  for _ in $(seq 1 30); do
    curl -fsS --max-time 3 http://127.0.0.1:8088/api/health | grep -q '"status":"ok"' && ok_api=1 && break
    sleep 2
  done
  [ "$ok_api" -eq 1 ] || {
    echo "错误: API 上游 127.0.0.1:8088/api/health 不可用" >&2
    return 1
  }

  code="$(curl -k -sS -o /dev/null -w '%{http_code}' --max-time 10 "https://${domain}" || true)"
  if [ "$code" = "502" ] || [ "$code" = "000" ]; then
    echo "检测到 https://${domain} 返回 ${code}，尝试自动重载宿主机 Nginx 后重试..."
    run_sudo systemctl reload nginx 2>/dev/null || true
    sleep 1
    code="$(curl -k -sS -o /dev/null -w '%{http_code}' --max-time 10 "https://${domain}" || true)"
  fi

  if [ "${code:-000}" -ge 500 ] || [ "${code:-000}" = "000" ]; then
    echo "错误: https://${domain} 返回 ${code}，部署后健康检查失败。" >&2
    echo "==== 诊断：80/443 监听 ====" >&2
    run_sudo ss -tlnp | sed -n '1p;/\:80 \|:443 /p' >&2 || true
    echo "==== 诊断：最近 Nginx 错误日志 ====" >&2
    run_sudo sh -c 'tail -n 80 /var/log/nginx/error.log 2>/dev/null || true' >&2
    return 1
  fi

  echo "健康检查通过：https://${domain} -> ${code}"
}