HAProxy 和后端服务器之间丢失一半的数据

我在 Ubuntu 上有这样的设置

浏览器 --> HAProxy --> 后端服务器

后端服务器是一个 ASP.NET Core Web 应用程序。

它的工作时间为 99.9%，除非正在上传二进制文件（带有多部分表单数据的简单 POST），在这种情况下我会收到错误消息：

System.IO.IOException: Unexpected end of Stream, the content may have already been read by another component. 

   at Microsoft.AspNetCore.WebUtilities.MultipartReaderStream.ReadAsync(Byte[] buffer, Int32 offset, Int32 count, CancellationToken cancellationToken)

   at Microsoft.AspNetCore.WebUtilities.FileBufferingReadStream.ReadAsync(Byte[] buffer, Int32 offset, Int32 count, CancellationToken cancellationToken)

   at Microsoft.AspNetCore.WebUtilities.StreamHelperExtensions.DrainAsync(Stream stream, ArrayPool`1 bytePool, Nullable`1 limit, CancellationToken cancellationToken)

   at Microsoft.AspNetCore.Http.Features.FormFeature.InnerReadFormAsync(CancellationToken cancellationToken)

   at Microsoft.AspNetCore.Antiforgery.DefaultAntiforgeryTokenStore.GetRequestTokensAsync(HttpContext httpContext)

   at Microsoft.AspNetCore.Antiforgery.DefaultAntiforgery.ValidateRequestAsync(HttpContext httpContext)

   at Microsoft.AspNetCore.Mvc.ViewFeatures.Filters.ValidateAntiforgeryTokenAuthorizationFilter.OnAuthorizationAsync(AuthorizationFilterContext context)

   at Microsoft.AspNetCore.Mvc.Infrastructure.ResourceInvoker.<InvokeFilterPipelineAsync>g__Awaited|19_0(ResourceInvoker invoker, Task lastTask, State next, Scope scope, Object state, Boolean isCompleted)

   at Microsoft.AspNetCore.Mvc.Infrastructure.ResourceInvoker.<InvokeAsync>g__Logged|17_1(ResourceInvoker invoker)

   at Microsoft.AspNetCore.Routing.EndpointMiddleware.<Invoke>g__AwaitRequestTask|6_0(Endpoint endpoint, Task requestTask, ILogger logger)

   at Microsoft.AspNetCore.Diagnostics.DeveloperExceptionPageMiddleware.Invoke(HttpContext context)

所以，我一直在尝试调查原因。

我直接尝试了 ASP.NET Core 后端，它可以工作。

所以我将传入的数据记录到后端，发现如果我直接这样做，我会得到预期的请求标头，然后是正文中的上传文件。

然后我通过 HAProxy 将传入的数据记录到后端，并且一半的数据丢失了。

IE。后端看到的第一个数据似乎是

04 E2 22 FC 60 FF 2B E1  BF 85 D2 75 F9 44 94 86

但这些字节大约是我文件的一半，即 25,126 字节。我根本看不到任何标题信息。

我准备接受我的日志记录不完善且不一定准确。似乎以某种方式将请求分成两部分，或者缓冲区已被填充然后被覆盖。

问题可能是什么？

global

    chroot /var/lib/haproxy
    log /dev/log    local0 info

    stats socket /run/haproxy/admin.sock mode 660 level admin
    stats timeout 30s
    user haproxy
    group users
    daemon
        
    lua-load /home/user/haproxy-mapping.lua

    ssl-default-bind-ciphers ECDH+AESGCM:ECDH+CHACHA20:DH+AESGCM:ECDH+AES256:DH+AES256:ECDH+AES128:DH+AES:RSA+AESGCM:RSA+AES:!aNULL:!MD5:!DSS:!AESCCM
    ssl-default-server-options no-sslv3 no-tlsv10 no-tlsv11 no-tls-tickets
    tune.ssl.default-dh-param 2048


defaults
    log     global
    mode    http

    timeout connect 30000
    timeout client  50000
    timeout server  50000
    option forwardfor
    option http-server-close #have also try http-keep-alive


frontend httpfront
    mode http
    bind *:80
    redirect scheme https code 301 if !{ ssl_fc }


frontend web_front_end

    bind *:443 ssl crt /home/.....file.pem
    mode http

     log /var/lib/haproxy/dev/log local0 info

        # Rate limiting
        stick-table  type ip  size 100k  expire 600s  store http_req_rate(60s) #store up to 100k requests for 60s, see if over 60s there are more than 600
        http-request track-sc0 src
        http-request deny deny_status 429 if { sc_http_req_rate(0) gt 600 }


    # Ensure we have a clean state to start with
    http-request del-header X-SERVER-SNI

    # Set the concatenated value of the SNI value to a temporary header
    http-request set-header X-SERVER-SNI haproxy.%[ssl_fc_sni] if { ssl_fc_sni -m found }

    # Set the value of the header to a transaction-level variable
    http-request set-var(txn.fc_sni) ssl_fc_sni #hdr(X-SERVER-SNI) if { hdr(X-SERVER-SNI) -m found }

    #use Lua code to determine which backend to send to
    use_backend %[lua.legacy_new_backend]





backend backendnodes_1_https
    balance roundrobin
    option forwardfor
    server node1 127.0.0.1:446 ssl verify none sni var(txn.fc_sni)