name: Recover DB (conditional mongo restart)

# 磁盘已腾出，但满盘期间 MongoDB 可能卡进坏状态。先用应用自带 driver ping：
#   卡死/失败 → 重启本机 mongod（共享库，恢复所有连库产品）；健康 → 只重启本产品 API 清连接池。
# 顺带装 pm2-logrotate，防止失控日志再次撑爆磁盘。
on:
  workflow_dispatch:

jobs:
  recover:
    runs-on: ubuntu-latest
    steps:
      - name: SSH recover DB + prevent log blowup
        uses: appleboy/ssh-action@v1
        with:
          host: ${{ secrets.ECS_HOST }}
          username: ${{ secrets.ECS_USER }}
          password: ${{ secrets.ECS_PASSWORD }}
          command_timeout: 10m
          script: |
            echo "========== 当前磁盘 =========="; df -h / | tail -1
            echo "========== 0) 装 pm2-logrotate 防再次爆盘(幂等) =========="
            pm2 install pm2-logrotate 2>&1 | tail -2 || echo "  (logrotate 安装跳过)"
            pm2 set pm2-logrotate:max_size 50M 2>/dev/null
            pm2 set pm2-logrotate:retain 3 2>/dev/null
            pm2 set pm2-logrotate:compress true 2>/dev/null
            echo "  pm2-logrotate: max_size=50M retain=3 compress=on"
            echo "========== 1) 用应用 driver ping MongoDB(8s 超时) =========="
            cd /var/www/digitalme-api.invoratec.cn
            PING=$(timeout 14 node -e '
              const { MongoClient } = require("mongodb");
              (async () => {
                const c = new MongoClient("mongodb://localhost:27017", { serverSelectionTimeoutMS: 8000 });
                try { await c.connect(); await c.db("admin").command({ ping: 1 }); console.log("MONGO_OK"); }
                catch (e) { console.log("MONGO_FAIL " + e.message); }
                finally { try { await c.close(); } catch (_) {} }
              })();
            ' 2>&1 | grep -aE "MONGO_OK|MONGO_FAIL" | head -1)
            echo "  ping 结果: [$PING]"
            if echo "$PING" | grep -q "MONGO_OK"; then
              echo "========== 2) Mongo 健康 → 只重启本产品 API 清连接池 =========="
              pm2 restart digitalme-api --update-env 2>/dev/null | tail -2
            else
              echo "========== 2) Mongo 卡死/无响应 → 重启本机 mongod =========="
              if systemctl list-units --type=service 2>/dev/null | grep -q mongod; then
                systemctl restart mongod 2>&1 | tail -3; sleep 8
                systemctl is-active mongod 2>/dev/null && echo "  mongod active" || echo "  ⚠ mongod 未 active"
              elif command -v docker >/dev/null && docker ps -a --format '{{.Names}}' 2>/dev/null | grep -qi mongo; then
                MC=$(docker ps -a --format '{{.Names}}' | grep -i mongo | head -1); echo "  docker restart $MC"; docker restart "$MC" 2>&1 | tail -2; sleep 8
              else
                echo "  ❌ 未识别 mongod 服务(既非 systemd 也非 docker)，需人工处理"
              fi
              echo "========== 3) 重启本产品 API =========="
              pm2 restart digitalme-api --update-env 2>/dev/null | tail -2
            fi
            sleep 8
            echo "========== 4) 验证 =========="
            curl -s -m 8  -o /dev/null -w "  /health     -> code=%{http_code} time=%{time_total}s\n" http://localhost:3603/health
            curl -s -m 12 -o /dev/null -w "  /api/plaza  -> code=%{http_code} time=%{time_total}s\n" http://localhost:3603/api/plaza
            curl -s -m 12 -o /dev/null -w "  /api/masters-> code=%{http_code} time=%{time_total}s\n" http://localhost:3603/api/masters
            echo "========== pm2 关键进程状态 =========="
            pm2 list 2>/dev/null | grep -iE "digitalme-api|mongo" || true
            echo "========== 最终磁盘 =========="; df -h / | tail -1
            echo "========== 完成 =========="