nginx服务器上的超时错误不一致

我有一个应用程序试图从Javascript应用程序的Ajax请求从nginx服务器检索xml文件。 通常情况下一切正常，但我经常在nginx日志中看到错误（并从我的应用程序错误报告中获取错误），nginx在获取期间显示超时：

2015/11/16 21:15:21 [error] 1208#0: *4894044 upstream timed out (110: Connection timed out) while connecting to upstream, client: 209.95.138.54, server: www.servername.com, request: "GET /Shape%20Textures/Metal/Born%20to%20Shine.jpg?agentView=436314 HTTP/1.1", upstream: "http://img.zgserver.com/nginx/Born to Shine.jpg?agentView=436314", host: "www.servername.com.com", referrer: "https://www.servername.com.com/?nid=39956&mode=edit" 

我们也有时会得到这个类似的错误：

 2015/11/17 19:03:16 [error] 1002#0: *54042 upstream timed out (110: Connection timed out) while reading response header from upstream, client: 137.164.121.52, server: www.servername.com, request: "POST /projects/?q=node/36375/update_project_time_and_stats HTTP/1.1", upstream: "http://127.0.0.1:3000/projects/?q=node/36375/update_project_time_and_stats", host: "www.servername.com", referrer: "https://www.servername.com/AQO/?nid=36375&mode=edit" 

我似乎有类似的超时类似的post，但所有这些post似乎可重现。 这个问题从来没有发生在我身上，但运行在现场服务器上，我会看到这些超时一天5-30。

这是我的nginxconfiguration：

• meteorMupx与Nginx反向代理
• 将Nginxconfiguration为Couchdb的反向代理
• 反向代理HTTP / 2从h2到h2c
• 反向代理：在Nginx上使用Bokeh服务器的Flask应用程序
• NGINX前面的Kibana没有使用“_plugin / kibana /”URI

 client_max_body_size 50M; server { server_name servername.com; return 301 $scheme://www.servername.com$request_uri; } server { listen 80; listen 443 ssl; fastcgi_read_timeout 120; ssl_certificate /path/to/ssl/star_servername_com.pem; ssl_certificate_key /path/to/ssl/star_servername_com.key; # Redirect all non-SSL traffic to SSL. if ($ssl_protocol = "") { rewrite ^ https://$host$request_uri? permanent; } root /usr/share/nginx/html; index index.html index.htm; # Make site accessible from http://localhost/ server_name www.servername.com; location / { proxy_set_header X-Forwarded-For $remote_addr; proxy_pass http://127.0.0.1:3000; proxy_read_timeout 120s; } } 

由于我不能复制它，我不知道我怎么能够追查这个问题？