.NET 6 Web API从Azure存储下载多份大文件Zip的CORS问题及优化咨询

我在.NET 6中开发了一个API，功能是从Azure存储下载任意数量的大文件并打包成Zip格式返回。API核心代码如下：

//Get the document name lists from database
List<MyDocuments> docLists = await GetMyDocumentNameAsync(countryIds);

string accessKey = "*******************************";
string containerName = "TestFolderName";

using (MemoryStream zipStream = new MemoryStream())
{
    using (ZipArchive zipArchive = new ZipArchive(zipStream, ZipArchiveMode.Create, true))
    {
        BlobServiceClient blobServiceClient = new BlobServiceClient(accessKey);
        BlobContainerClient containerClient = blobServiceClient.GetBlobContainerClient(containerName);

        foreach (var doc in docLists)
        {
            BlobClient blobClient = containerClient.GetBlobClient(doc.FileURL);
            BlobDownloadInfo downloadInfo = await blobClient.DownloadAsync();

            ZipArchiveEntry entry = zipArchive.CreateEntry(doc.FileName);
            using (Stream entryStream = entry.Open())
            {
                await downloadInfo.Content.CopyToAsync(entryStream);
            }
        }
    }

    // Reset the stream position to the beginning
    zipStream.Position = 0;
    
    return File(zipStream, "application/zip", "ZipFile.zip");
}

从Angular应用调用该API时遇到CORS错误，已在Startup.cs配置CORS策略：

services.AddCors(options => options.AddPolicy("CorsPolicy",
    builder => builder
        .WithOrigins(_appConfiguration["MyCors"].Split(",").Select(o =>o.RemovePostFix("/")).ToArray())
        .AllowAnyHeader()
        .AllowAnyMethod()
        .AllowCredentials())
);

Appsettings配置：

"MyCors": "http://localhost,http://localhost:4200"

报错信息：

• 截图1：请求的资源上不存在'Access-Control-Allow-Origin'标头
• 截图2：预检请求的响应未通过访问控制检查

1. 确认CORS策略已正确启用
   仅注册策略还不够，需在Startup.cs的Configure方法中添加中间件启用，注意顺序要在UseRouting之后、UseAuthorization之前：

   app.UseCors("CorsPolicy");
   

2. 检查Origin配置格式

   • 确保MyCors中的地址无多余空格（如http://localhost,http://localhost:4200不要写成http://localhost, http://localhost:4200）
   • 核对Angular应用实际运行地址，若使用HTTPS启动，需将配置改为https://localhost:4200

3. 处理预检请求与响应头

   • 若请求包含自定义头或复杂操作（如带Credentials），需确保服务器正确响应OPTIONS请求（AllowAnyMethod已包含OPTIONS，若仍有问题可显式声明）
   • 添加暴露响应头配置，允许客户端获取Content-Disposition等必要头：

     builder.WithExposedHeaders("Content-Disposition");
     

4. 优化响应流返回方式
   替换File方法为FileStreamResult，确保响应头正确设置：

   return new FileStreamResult(zipStream, "application/zip")
   {
       FileDownloadName = "ZipFile.zip"
   };
   

当前代码将所有文件缓冲到MemoryStream，大文件/多文件场景下会导致内存占用过高，以下是优化方案：

1. 边下载边打包边返回，避免内存缓冲
   直接将Zip内容写入API响应流，减少内存占用：

   var response = HttpContext.Response;
   response.Headers.Add("Content-Disposition", "attachment; filename=ZipFile.zip");
   response.ContentType = "application/zip";
   
   using (ZipArchive zipArchive = new ZipArchive(response.Body, ZipArchiveMode.Create, true))
   {
       BlobServiceClient blobServiceClient = new BlobServiceClient(accessKey);
       BlobContainerClient containerClient = blobServiceClient.GetBlobContainerClient(containerName);
   
       foreach (var doc in docLists)
       {
           BlobClient blobClient = containerClient.GetBlobClient(doc.FileURL);
           using (var blobStream = await blobClient.OpenReadAsync())
           {
               ZipArchiveEntry entry = zipArchive.CreateEntry(doc.FileName);
               using (Stream entryStream = entry.Open())
               {
                   await blobStream.CopyToAsync(entryStream, 81920); // 80KB缓冲区提升拷贝效率
               }
           }
       }
   }
   
   return new EmptyResult();
   

2. 分块下载单个超大文件
   对超过1GB的文件，采用分块下载避免一次性加载整个文件到内存：

   var blobProperties = await blobClient.GetPropertiesAsync();
   long fileSize = blobProperties.Value.ContentLength;
   long offset = 0;
   const long chunkSize = 1024 * 1024 * 8; // 8MB分块
   
   using (Stream entryStream = entry.Open())
   {
       while (offset < fileSize)
       {
           long currentChunkSize = Math.Min(chunkSize, fileSize - offset);
           BlobDownloadInfo downloadInfo = await blobClient.DownloadAsync(new HttpRange(offset, currentChunkSize));
           await downloadInfo.Content.CopyToAsync(entryStream);
           offset += currentChunkSize;
       }
   }
   

3. 可控并行下载多文件
   针对多文件场景，限制并发数避免超出Azure存储请求限制：

   var semaphore = new SemaphoreSlim(5); // 限制并发数为5
   var tasks = docLists.Select(async doc =>
   {
       await semaphore.WaitAsync();
       try
       {
           BlobClient blobClient = containerClient.GetBlobClient(doc.FileURL);
           using (var blobStream = await blobClient.OpenReadAsync())
           {
               ZipArchiveEntry entry = zipArchive.CreateEntry(doc.FileName);
               using (Stream entryStream = entry.Open())
               {
                   await blobStream.CopyToAsync(entryStream, 81920);
               }
           }
       }
       finally
       {
           semaphore.Release();
       }
   });
   
   await Task.WhenAll(tasks);
   

4. 预处理文件减轻API压力
   若频繁打包大量超大文件，可将打包逻辑移至Azure函数，提前生成Zip文件存储到Blob，API仅返回下载链接。

内容的提问来源于stack exchange，提问作者mnu-nasir