通过对等机进行httputil.ReverseProxy连接重置

我的任务是制作一个反向代理应用程序，从请求/响应正文中提取一些特定信息，并将其发送到logstash。所以我决定先做一些负载测试。我编写了非常简单的后端web服务器和基本代理

后端服务器代码：

反向代理服务器代码：

正如您所看到的，后端服务器正在侦听

localhost:3001

和代理侦听

localhost:3000

我正在使用ApacheBenchmark进行负载测试

当我运行

时$ab-c20-n20000-s10http://127.0.0.1:3000/

大约12k-14k请求代理开始写入错误，如：

2016/12/23 13:02:50 http: proxy error: read tcp [::1]:58330->[::1]:3001: read: connection reset by peer
2016/12/23 13:02:50 http: proxy error: read tcp [::1]:58461->[::1]:3001: read: connection reset by peer
2016/12/23 13:02:50 http: proxy error: read tcp [::1]:58977->[::1]:3001: read: connection reset by peer
2016/12/23 13:02:50 http: proxy error: read tcp [::1]:59459->[::1]:3001: read: connection reset by peer
...

不久之后，阿帕奇板凳超时死亡。一些试验结果：

Benchmarking 127.0.0.1 (be patient)
Completed 2000 requests
Completed 4000 requests
Completed 6000 requests
Completed 8000 requests
Completed 10000 requests
^C

Server Software:        
Server Hostname:        127.0.0.1
Server Port:            3000

Document Path:          /
Document Length:        4 bytes

Concurrency Level:      20
Time taken for tests:   4.868 seconds
Complete requests:      11514
Failed requests:        22
   (Connect: 0, Receive: 0, Length: 22, Exceptions: 0)
Non-2xx responses:      22
Total transferred:      1381790 bytes
HTML transferred:       45968 bytes
Requests per second:    2365.28 [#/sec] (mean)
Time per request:       8.456 [ms] (mean)
Time per request:       0.423 [ms] (mean, across all concurrent requests)
Transfer rate:          277.20 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    1   4.2      1     111
Processing:     0    3   4.8      3     113
Waiting:        0    3   4.7      3     113
Total:          0    4   6.4      4     116

Percentage of the requests served within a certain time (ms)
  50%      4
  66%      4
  75%      5
  80%      5
  90%      5
  95%      6
  98%      6
  99%      7
 100%    116 (longest request)

还有更多，如果我在第一次测试结束后立即再次启动

ab

，我将收到如下信息：

Benchmarking 127.0.0.1 (be patient)
apr_pollset_poll: The timeout specified has expired (70007)
Total of 52 requests completed

（可能是之前测试中的某些连接仍然有效）

互联网称Go中有一些DoS保护机制，显然这是连接重置的原因。 所以我的问题是：

重置连接的原因（读取缓冲区已满、发生超时、连接太多或可能我的代码已断开）

我可以调整这个机制吗？例如，扩展更改最大超时的缓冲区

一些提示说明如何正确处理此类错误

---

谢谢。

ab

正在使用http/1.0而不使用keepalive。不要使用

ab

。ab不是你的最佳选择。看看wrk，它像ab，但在很多方面都更好，或者vegeta（）可能不是最好的，我肯定会尝试其他解决方案（谢谢），但我不能保证这个问题不会在生产中发生。所以我现在想知道如何修复/调试它。@MarkBelotskiy:see，和@JimB，这样它可以是OS限制或http.Transport限制。。。谢谢你的提示。我两个都试试，然后报告。

Benchmarking 127.0.0.1 (be patient)
Completed 2000 requests
Completed 4000 requests
Completed 6000 requests
Completed 8000 requests
Completed 10000 requests
^C

Server Software:        
Server Hostname:        127.0.0.1
Server Port:            3000

Document Path:          /
Document Length:        4 bytes

Concurrency Level:      20
Time taken for tests:   4.868 seconds
Complete requests:      11514
Failed requests:        22
   (Connect: 0, Receive: 0, Length: 22, Exceptions: 0)
Non-2xx responses:      22
Total transferred:      1381790 bytes
HTML transferred:       45968 bytes
Requests per second:    2365.28 [#/sec] (mean)
Time per request:       8.456 [ms] (mean)
Time per request:       0.423 [ms] (mean, across all concurrent requests)
Transfer rate:          277.20 [Kbytes/sec] received

Connection Times (ms)
              min  mean[+/-sd] median   max
Connect:        0    1   4.2      1     111
Processing:     0    3   4.8      3     113
Waiting:        0    3   4.7      3     113
Total:          0    4   6.4      4     116

Percentage of the requests served within a certain time (ms)
  50%      4
  66%      4
  75%      5
  80%      5
  90%      5
  95%      6
  98%      6
  99%      7
 100%    116 (longest request)

Benchmarking 127.0.0.1 (be patient)
apr_pollset_poll: The timeout specified has expired (70007)
Total of 52 requests completed