uwsgi性能调忧
摘要：调大uwsgi配置中 listen=1024的数目是提高并发能力最有效的办法。第二种方法是调大processes数目
发现总是超时报警
1 使用ab确定网络具体征况
[bre@dmp-1 ~]$ ab -n 1000 -c 100 http://.com/Mapping.do?
Concurrency Level: 100
Time taken for tests: 1.040 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Non-2xx responses: 1000
Total transferred: 486000 bytes
HTML transferred: 0 bytes
Requests per second: 961.42 [#/sec] (mean)
Time per request: 104.013 [ms] (mean)
Time per request: 1.040 [ms] (mean, across all concurrent requests)
Transfer rate: 456.30 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.4 0 2
Processing: 3 28 61.9 25 1005
Waiting: 3 28 61.9 25 1005
Total: 5 28 61.9 25 1006
Percentage of the requests served within a certain time (ms)
50% 25
66% 26
75% 26
80% 26
90% 27
95% 27
98% 29
99% 31
100% 1006 (longest request)
[bre@dmp-1 ~]$ ab -n 1000 -c 200 http://.com/Mapping.do?
Concurrency Level: 200
Time taken for tests: 3.022 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Non-2xx responses: 909
Total transferred: 463432 bytes
HTML transferred: 0 bytes
Requests per second: 330.91 [#/sec] (mean)
Time per request: 604.403 [ms] (mean)
Time per request: 3.022 [ms] (mean, across all concurrent requests)
Transfer rate: 149.76 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 1 1.2 0 4
Processing: 6 274 791.9 28 3015
Waiting: 6 274 791.9 28 3015
Total: 10 275 792.5 29 3018
Percentage of the requests served within a certain time (ms)
50% 29
66% 29
75% 30
80% 30
90% 43
95% 3015
98% 3015
99% 3016
100% 3018 (longest request)
[bre@dmp-1 ~]$ ab -n 1000 -c 300 http://.com/Mapping.do?
Concurrency Level: 300
Time taken for tests: 3.036 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Non-2xx responses: 811
Total transferred: 439128 bytes
HTML transferred: 0 bytes
Requests per second: 329.35 [#/sec] (mean)
Time per request: 910.892 [ms] (mean)
Time per request: 3.036 [ms] (mean, across all concurrent requests)
Transfer rate: 141.24 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 2 2.4 0 6
Processing: 5 521 1043.3 29 3027
Waiting: 5 521 1043.3 29 3027
Total: 10 523 1044.8 29 3032
Percentage of the requests served within a certain time (ms)
50% 29
66% 31
75% 35
80% 53
90% 3021
95% 3027
98% 3029
99% 3030
100% 3032 (longest request)
[bre@dmp-1 ~]$
[bre@dmp-1 ~]$ ab -n 1000 -c 400 http://.com/Mapping.do?
Concurrency Level: 400
Time taken for tests: 3.025 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Non-2xx responses: 711
Total transferred: 414328 bytes
HTML transferred: 0 bytes
Requests per second: 330.58 [#/sec] (mean)
Time per request: 1209.982 [ms] (mean)
Time per request: 3.025 [ms] (mean, across all concurrent requests)
Transfer rate: 133.76 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 3 3.7 0 9
Processing: 6 738 1148.8 28 3013
Waiting: 6 738 1148.8 28 3013
Total: 13 741 1151.3 28 3022
Percentage of the requests served within a certain time (ms)
50% 28
66% 38
75% 2011
80% 2016
90% 3013
95% 3016
98% 3019
99% 3019
100% 3022 (longest request)
[bre@dmp-1 ~]$ ab -n 1000 -c 500 http://cm.api.baifendian.com/Mapping.do?bfd_nid=behe
This is ApacheBench, Version 2.3 <$Revision: 655654 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/
Benchmarking cm.api.baifendian.com (be patient)
Completed 100 requests
Completed 200 requests
Completed 300 requests
Completed 400 requests
Completed 500 requests
Completed 600 requests
Completed 700 requests
Completed 800 requests
Completed 900 requests
Completed 1000 requests
Finished 1000 requests
Server Software: nginx/1.2.9
Server Hostname: cm.api.baifendian.com
Server Port: 80
Document Path: /Mapping.do?bfd_nid=behe
Document Length: 0 bytes
Concurrency Level: 500
Time taken for tests: 3.100 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Non-2xx responses: 610
Total transferred: 389280 bytes
HTML transferred: 0 bytes
Requests per second: 322.63 [#/sec] (mean)
Time per request: 1549.754 [ms] (mean)
Time per request: 3.100 [ms] (mean, across all concurrent requests)
Transfer rate: 122.65 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 5 4.7 9 10
Processing: 9 930 1168.2 31 3014
Waiting: 9 930 1168.2 31 3014
Total: 18 935 1171.1 33 3023
Percentage of the requests served within a certain time (ms)
50% 33
66% 2012
75% 2016
80% 2018
90% 3018
95% 3020
98% 3021
99% 3021
100% 3023 (longest request)
[bre@dmp-1 ~]$ ab -n 1000 -c 600 http://cm.api.baifendian.com/Mapping.do?bfd_nid=behe
This is ApacheBench, Version 2.3 <$Revision: 655654 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/
Benchmarking cm.api.baifendian.com (be patient)
Completed 100 requests
Completed 200 requests
Completed 300 requests
Completed 400 requests
Completed 500 requests
Completed 600 requests
Completed 700 requests
Completed 800 requests
Completed 900 requests
Completed 1000 requests
Finished 1000 requests
Server Software: nginx/1.2.9
Server Hostname: cm.api.baifendian.com
Server Port: 80
Document Path: /Mapping.do?bfd_nid=behe
Document Length: 0 bytes
Concurrency Level: 600
Time taken for tests: 3.039 seconds
Complete requests: 1000
Failed requests: 0
Write errors: 0
Non-2xx responses: 511
Total transferred: 364728 bytes
HTML transferred: 0 bytes
Requests per second: 329.08 [#/sec] (mean)
Time per request: 1823.243 [ms] (mean)
Time per request: 3.039 [ms] (mean, across all concurrent requests)
Transfer rate: 117.21 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 7 5.2 10 12
Processing: 9 1126 1170.6 45 3022
Waiting: 9 1126 1170.6 45 3022
Total: 21 1133 1173.4 55 3033
Percentage of the requests served within a certain time (ms)
50% 55
66% 2014
75% 2017
80% 2019
90% 3019
95% 3025
98% 3028
99% 3031
100% 3033 (longest request)
可以看到，每秒处理请求数目极限大约是300，随着并发数目的增大，性能开始下降。
现在的请求每天1000000条，那么高峰时段的请求次数为：1000000/（8*60*60）*10 = 347 （次每秒）
每天正常请求时段是8小时，高峰时段请求频次是平均的8～10倍。
所以只是勉强够用而已，还需要提高速度（优化整个程序流程或者增大并发进程）。
2 使用 uwsgitop 查看是前端（nginx）还是后端（uwsgi和后续处理流程）
观察结果
uwsgi-1.9.14 - Mon Feb 10 16:17:16 2014 - req: 188802 - lq: 0 - tx: 66.0M node: dmp-1 - cwd: /opt/bre/CookieMapping/dsWebService - uid: 508 - gid: 503 - masterpid: 25062
WID % PID REQ EXC SIG STATUS AVG RSS VSZ TX RunT
7 41.4 25073 78127 0 0 idle 1ms 0 0 25.0M 217637
5 10.0 25071 18956 0 0 idle 1ms 0 0 6.0M 105123
8 9.1 25076 17213 0 0 idle 1ms 0 0 6.0M 101432
4 9.1 25070 17196 0 0 idle 1ms 0 0 6.0M 99265
6 9.0 25072 17032 0 0 idle 1ms 0 0 6.0M 95653
2 7.6 25068 14312 0 0 idle 1ms 0 0 5.0M 93229
1 7.1 25067 13489 0 0 idle 17ms 0 0 5.0M 85652
3 6.6 25069 12477 0 0 idle 2ms 0 0 4.0M 91285
发现当使用ab测试时，当并发比较大时候，时间会出现很大的情况
3 措施
（1）首先，uwsgi观察到上面的负载不均衡，第7个线程负责的请求数远大于其他的。
调整使之均衡
（2）再机器能力足够的情况下，增加线程数目
（3）建立uwsgi集群，在另外一台机器上也部署此服务，并让nginx转发部分到此
uwsgi性能调忧的更多相关文章
- centos 7上nginx+uwsgi 性能调优
上一章将nginx +uwsgi搭建起来,将keystone挂载后面.但是发现一个问题,如果http请求达到一定量后,nginx直接返回502.这让需要部署大规模openstack集群的我很是头疼,比 ...
- web前端性能调优
最近2个月一直在做手机端和电视端开发,开发的过程遇到过各种坑.弄到快元旦了,终于把上线了.2个月干下来满满的的辛苦,没有那么忙了自己准备把前端的性能调优总结以下,以方便以后自己再次使用到的时候得于得心 ...
- ［网站性能2］Asp.net平台下网站性能调优的实战方案
文章来源:http://www.cnblogs.com/dingjie08/archive/2009/11/10/1599929.html 前言 最近帮朋友运营的平台进行了性能调优,效果还不错, ...
- Asp.net平台下网站性能调优的实战方案（转）
转载地址:http://www.cnblogs.com/chenkai/archive/2009/11/07/1597795.html 前言 最近帮朋友运营的平台进行了性能调优,效果还不错,所以写出来 ...
- 第0/24周 SQL Server 性能调优培训引言
大家好,这是我在博客园写的第一篇博文,之所以要开这个博客,是我对MS SQL技术学习的一个兴趣记录. 作为计算机专业毕业的人,自己对技术的掌握总是觉得很肤浅,博而不专,到现在我才发现自己的兴趣所在,于 ...
- sqlserver性能调优第一步
相信不少的朋友,无论是做开发.架构的,还是DBA等,都经常听说“调优”这个词.说起“调优”,可能会让很多技术人员心头激情澎湃,也可能会让很多人感觉苦恼,不知道如何入手.当然,也有很多人对此不屑一顾,因 ...
- JavaScript：内存泄露、性能调优
1.在进行JS内存泄露检查之前,先要了解JS的内存管理: https://developer.mozilla.org/en-US/docs/Web/JavaScript/Memory_Manageme ...
- hadoop 性能调优与运维
hadoop 性能调优与运维 . 硬件选择 . 操作系统调优与jvm调优 . hadoop运维 硬件选择 1) hadoop运行环境 2) 原则一: 主节点可靠性要好于从节点 原则二:多路多核,高频 ...
- JVM性能调优监控工具jps、jstack、jmap、jhat、jstat、hprof使用详解
摘要: JDK本身提供了很多方便的JVM性能调优监控工具,除了集成式的VisualVM和jConsole外,还有jps.jstack.jmap.jhat.jstat.hprof等小巧的工具,本博客希望 ...
随机推荐
- rndc 错误解决 和 远程配置
dc: connect failed: connection refusedrndc: connect failed: connection refused 解决办法:默认安装BIND9以后,是无法直 ...
- linux 进程命令
Linux中的ps命令是Process Status的缩写.ps命令用来列出系统中当前运行的那些进程.ps命令列出的是当前那些进程的快照,就是执行ps命令的那个时刻的那些进程,如果想要动态的显示进程信 ...
- 判断Android应用是否安装、运行
本文介绍3个方法,判断手机上是否安装了某应用.该应用是否正在运行.获取手机某个APK文件的安装Intent.启动该Intent就可以直接安装该APK. /** * 判断应用是否已安装 * * @par ...
- 留言本，keyCode
<!DOCTYPE HTML> <html> <head> <meta http-equiv="Content-Type" content ...
- play1.x vs play2.x 对比（转）
个人看到对比play1.x和play2.x比较的文章中,写的最深入,最清晰的一个.转自:http://freewind.me/blog/20120728/965.html 为了方便群中的Play初学者 ...
- UVALive 6885 Flowery Trails
两次SPFA #include<cstdio> #include<cstring> #include<cmath> #include<vector> # ...
- Win7与Ubuntu双系统时卸载Ubuntu的方法
Win7与Ubuntu双系统时卸载Ubuntu的方法 [日期:2010-03-26] 来源:Ubuntu社区 作者:Ubuntu编辑 [字体:大 中 小] 1. 下载MBRFix工具,放 ...
- L5,no wrong numbers
expressions: up to now,到现在为止 a great many,数量很大 in a way,在某种意义上说 words: burn,vt燃烧,vi烧毁,n灼烧 obtain,v ...
- java设计原则：16种原则
一 类的设计原则 1 依赖倒置原则-Dependency Inversion Principle (DIP) 2 里氏替换原则-Liskov Substitution Principle (L ...
- AutoFac使用方法总结
AutoFac是.net平台下的IOC容器产品,它可以管理类之间的复杂的依赖关系.在使用方面主要是register和resolve两类操作. 这篇文章用单元测试的形式列举了AutoFac的常用使用方法 ...