NextJS API路由本地正常，Vercel部署后数据插入异常求助

我的NextJS API路由在本地运行完全正常，调用后可在本地数据库中创建200+条记录，但部署到Vercel后出现异常：插入的记录数完全随机，有时0条、有时3条、有时55条。相关API路由代码如下：

import type { NextApiRequest, NextApiResponse } from "next";
import axios from "axios";
import prisma from "../../../lib/prisma";

interface Platform {
  id: number;
  abbreviation: string;
  alternative_name: string;
  category: number;
  created_at: number;
  generation: number;
  name: string;
  platform_logo: {
    id: number;
    alpha_channel: boolean;
    animated: boolean;
    height: number;
    image_id: string;
    url: string;
    width: number;
    checksum: string;
  };
  slug: string;
  summary: string;
  updated_at: number;
  url: string;
  versions: number[];
  checksum: string;
}

export default async function handler(
  req: NextApiRequest,
  res: NextApiResponse
) {
  const url = "https://api.igdb.com/v4/platforms";

  const headers = {
    "client-id": process.env.TWITCH_IGDB_CLIENT_ID,
    authorization: "Bearer my_token",
    "accept-encoding": "*",
  };

  const body = `fields *, platform_family.*, platform_logo.*;
  limit 500;`;

  const response = await axios.post(url, body, { headers });

  console.log("The response status is:", response.status);

  if (response.status == 200) {
    await response.data.forEach(async (p: Platform) => {
      try {
        const platform = await prisma.platform
          .create({
            data: {
              category: p.category,
              name: p.name,
              slug: p.slug,
              platform_logo_url: p.platform_logo?.url, // Todo, sanitize this url.
              hashtag: p.name, // Todo, convert name to hashtag
            },
          });

        console.log("Created platforms", platform);
      } catch (e) {
        console.log(e)
      }

    });
  } else {
    console.log(response.status);
  }

  res.status(200).json({ message: "Imported all platforms." });
}

问题根源在于**forEach不支持异步函数的等待机制**：

• forEach会遍历数组并立即触发每个异步插入操作，但它不会等待这些异步操作完成，而是直接继续执行后续代码。
• 本地环境中，Node.js进程会持续运行直到所有异步任务结束，所以所有插入都能完成。
• 但Vercel的Serverless函数有严格的生命周期：当handler函数执行到res.status(200).json(...)并返回响应后，函数进程会被立即终止，此时forEach中未完成的异步插入操作会被强制中断，导致插入记录数随机。

将forEach替换为Promise.all，确保所有异步插入操作全部完成后再返回响应：

import type { NextApiRequest, NextApiResponse } from "next";
import axios from "axios";
import prisma from "../../../lib/prisma";

interface Platform {
  id: number;
  abbreviation: string;
  alternative_name: string;
  category: number;
  created_at: number;
  generation: number;
  name: string;
  platform_logo: {
    id: number;
    alpha_channel: boolean;
    animated: boolean;
    height: number;
    image_id: string;
    url: string;
    width: number;
    checksum: string;
  };
  slug: string;
  summary: string;
  updated_at: number;
  url: string;
  versions: number[];
  checksum: string;
}

export default async function handler(
  req: NextApiRequest,
  res: NextApiResponse
) {
  const url = "https://api.igdb.com/v4/platforms";

  const headers = {
    "client-id": process.env.TWITCH_IGDB_CLIENT_ID,
    authorization: "Bearer my_token",
    "accept-encoding": "*",
  };

  const body = `fields *, platform_family.*, platform_logo.*;
  limit 500;`;

  const response = await axios.post(url, body, { headers });

  console.log("The response status is:", response.status);

  if (response.status == 200) {
    // 将forEach替换为Promise.all，等待所有插入操作完成
    await Promise.all(response.data.map(async (p: Platform) => {
      try {
        const platform = await prisma.platform.create({
          data: {
            category: p.category,
            name: p.name,
            slug: p.slug,
            platform_logo_url: p.platform_logo?.url,
            hashtag: p.name,
          },
        });
        console.log("Created platform", platform);
      } catch (e) {
        console.error("Failed to create platform:", e);
      }
    }));
  } else {
    console.error("IGDB API response status:", response.status);
    return res.status(response.status).json({ error: "Failed to fetch platforms" });
  }

  res.status(200).json({ message: "Imported all platforms." });
}

额外优化建议：

• 把硬编码的"Bearer my_token"替换为环境变量，比如process.env.TWITCH_IGDB_ACCESS_TOKEN，避免敏感信息泄露。
• 可以考虑使用批量插入（如Prisma的createMany）来减少数据库请求次数，提升效率并降低超时风险。

内容的提问来源于stack exchange，提问作者Only Bolivian Here