1. 问题现象
nginx运行过程中，发现无法访问nginx（导致nagios监控告警） ,且error日志中一直报如下错误
|
2011/05/10 16:31:07 [alert] 24082#0: 1024 worker_connections is not enough while accepting new connection on
0.0.0.0:81
2011/05/10 16:31:07 [alert] 24082#0: 1024 worker_connections is not enough while accepting new connection on
0.0.0.0:81
2011/05/10 16:31:07 [alert] 24085#0: 1024 worker_connections is not enough while accepting new connection on
0.0.0.0:81
2011/05/10 16:31:07 [alert] 24082#0: 1024 worker_connections is not enough while accepting new connection on
0.0.0.0:81
2011/05/10 16:31:07 [alert] 24085#0: 1024 worker_connections is not enough while accepting new connection on
0.0.0.0:81
2011/05/10 16:31:08 [alert] 24082#0: 1024 worker_connections is not enough while accepting new connection on
0.0.0.0:81
2011/05/10 16:31:08 [alert] 24082#0: 1024 worker_connections is not enough while accepting new connection on
0.0.0.0:81
2011/05/10 16:31:08 [alert] 24085#0: 1024 worker_connections is not enough while accepting new connection on
0.0.0.0:81
2011/05/10 16:31:08 [alert] 24082#0: 1024 worker_connections is not enough while accepting new connection on
0.0.0.0:81
|
2.初步原因分析
目前nginx 中worker_connections 值设置是1024，worker_processes 值设置是4，按反向代理模式下最大连接数的理论计算公式：
最大连接数 = worker_processes * worker_connections/4
算出目前每台nginx理论上只支持1024个。
但观察NGINX 历史性能数据，发现活动连接数经常是大于1024，并且最高达到5000左右，因而有时会导致连接数不够，直到NGINX无法再处理新的连接请求。
3.解决办法
查看相关资料，生产环境中worker_connections 建议值最好超过9000，计划将一台nginx 设置为10240，再观察一段时间。