uwsgi 异常服务器内存 cpu 爆满优化思路

> 本文由 [简悦 SimpRead](http://ksria.com/simpread/) 转码， 原文地址 [www.cnblogs.com](https://www.cnblogs.com/pyyu/p/11969133.html)

记录线上服务器通过 linux 性能检测工具 glances 检测到 cpu、内存爆满，且是 uwsgi 进程占用，对于服务器内核，以及 uwsgi 配置优化
===============================================================================

参考文章

```
https://blog.csdn.net/orangleliu/article/details/48531759

```

uwsgi.log 日志报错

```
Tue Jun  2 17:33:27 2015 - *** uWSGI listen queue of socket "127.0.0.1:9080" (fd: 3) full !!! (101/100) ***
Tue Jun  2 17:33:28 2015 - *** uWSGI listen queue of socket "127.0.0.1:9080" (fd: 3) full !!! (101/100) ***

```

解决思路：

1. 修改内核配置

2. 修改 uwsgi 配置

内核修改
----

```
#对于一个经常处理新连接的高负载 web服务环境来说，默认的 128 太小了
net.core.somaxconn = 262144
#表示SYN队列的长度，默认为1024，加大队列长度为8192，可以容纳更多等待连接的网络连接数
net.ipv4.tcp_max_syn_backlog = 8192
#网卡设备将请求放入队列的长度
net.core.netdev_max_backlog = 65536

```

修改完成之后要记得 `sysctl -p` 重新加载内核参数

uwsgi 修改
--------

修改 uwsgi.ini

```
listen=1024

```

重启 uwsgi

新版 uwsgi 配置文件，着实好用
==================

```
# 缓存体系
cache = true
cache2 = name=mycache,items=100
lazy-apps = true
wsgi-disable-file-wrapper = true
memory-report = true
uid = www
master = true
vhost = true
no-stie = true
workers = 10
reload-mercy = 10
vacuum = true
max-requests = 3500
limit-as = 1024
buffer-size = 65535
pidfile = /var/run/uwsgi.pid
master-fifo = /tmp/uwsgififo
#daemonize = /var/log/logs/uwsgi.log
logto = /var/log/logs/uwsgi.log
log-maxsize = 10000000000
enable-threads=true
die-on-term=true
wsgi-disable-file-wrapper=true
env= LUFFY_ENV=Production
ksm=10
ignore-sigpipe = true
ignore-write-errors = true
disable-write-exception = true
harakiri = 30
thunder-lock = True

```