SRE文档

深浅模式

104 案例 ​

查看nginx error错误，发现上传接口报以下错：

2019/10/10 19:58:25 [error] 299784#0: *5967188 readv() failed (104: Connection reset by peer) while reading upstream, client: 59.34.155.7, server: xxxxxxxx, request: "POST /stream/tracking/file HTTP/1.1", upstream: "http://xxxxxxxx/stream/tracking/file", host: "xxxxxxxx"

这种错误日志不多，第一感觉就是上传文件过大，传输时间过长，然后连接被中断。

当使用nginx作为反向代理时，为了支持长连接，需要做到两点：

从client到nginx的连接是长连接，对于客户端来说，nginx长连接是默认开启的。从nginx到server的连接是长连接，需要自己开启

bash
upstream bigdata {
    server 10.0.20.xx:18018;
    server 10.0.20.xx:18018;
    server 10.0.20.xx:18018;
    server 10.0.20.xx:18018;
    keepalive 100;   //根据qps来调整
}  

location ~ / {
	proxy_pass http://bigdata;
    #下面的timeout跟自己的业务相关设置对应的timeout
    proxy_connect_timeout      120;   //加大120
    proxy_send_timeout         120;   //加大120
    proxy_read_timeout         120;   //加大120
    proxy_http_version 1.1;    //开启后端，长连接
    proxy_set_header Connection "";  //开启后端，长连接
}
upstream bigdata {
    server 10.0.20.xx:18018;
    server 10.0.20.xx:18018;
    server 10.0.20.xx:18018;
    server 10.0.20.xx:18018;
    keepalive 100;   //根据qps来调整
}  

location ~ / {
	proxy_pass http://bigdata;
    #下面的timeout跟自己的业务相关设置对应的timeout
    proxy_connect_timeout      120;   //加大120
    proxy_send_timeout         120;   //加大120
    proxy_read_timeout         120;   //加大120
    proxy_http_version 1.1;    //开启后端，长连接
    proxy_set_header Connection "";  //开启后端，长连接
}

注意：keepalive指定的数值是Nginx每个worker连接后端的最大长连接数，而不是整个Nginx的

https://blog.csdn.net/zzhongcy/article/details/89090193

https://www.cnblogs.com/yehaifeng/p/11819241.html

http://blog.51yip.com/category/safe