Battle test(Locust and Apache Bench)

1.目的

WEB负载测试(Load testing)可用于评估WEB系统的最大可服务容量，同时识别系统中潜在的性能瓶颈,
提高系统在正常以及接近极限访问量下的可靠性。
负载测试中，当负载不断增加,超过被测系统的容量上限时，此时变成了压力测试（stress testing)，被测试系统一般会逐渐服务不正常，但是系统行为需要在预期的范围，待压力测试持续一段时间后（如1小时），负载变轻时，系统应能能自动恢复正常（不需要人工干预）。

负载测试技术指标

• Requests per second：
  每秒能支持的请求数

• Concurrent users：
  能支持的用户并发请求数(同时发送请求的用户数)

• Response time per request：
  单次请求的响应时间（包括平均，最大，标准差）

• Throughput：
  吞吐量，表示每秒能支持的数据下载访问量，侧重衡量网络性能

• Availability：
  可用性，一段时间的请求成功数/总请求数。

tools

1.Locust

What is Locust:
Locust is an easy-to-use,distributed, user load testing tool. It s is intended for load-testing web sites(or other systems) and figuring out how many concurrent users a
system can handle.

The idea is that during a test, a swarm of locusts will attack your websiute. The behavior of each locust is defined by you and the swarming process is monitored from a web UI in real-time. This will help you battle test and identify bottllenecks in your code before letting real users in.

Links

• github:    https://github.com/locustio/locust
• Website:   http://locust.io/
• Documentation:   http://docs.locust.io/en/latest/api.html
• Support/Questions:   https://slack.locust.io/

How to use?

No need for clunky UIs or bloated XML - just code as you normally would(python programming). Based of coroutines instead of callbacks, your code looks and behaves like normal ,blocking python code.

Demo1:Hello world

#demo.py
from locust import Locust,TaskSet, task                                                                                                                                        


class MyTaskSet(TaskSet):
    @task
    def my_task(self):
        print "executing my task"



class MyLocust(Locust):
    task_set = MyTaskSet
    min_wait = 5000
    max_wait = 12000

How to start?

when in linux terminal,input:
locust -f demo.py –host = http://www.baidu.com (or other url)
When you have started Locust using the above command liens, you should open up a browser and point to http://127.0.0.1:8089(if Locust in local).Then you should filled 2 things: Number of users to simulate and Hatch rate.

---

Demo2:from official website(login before hack)

#locustfile.py
from locust import HttpLocust, TaskSet

def login(l):
    l.client.post("/login", {"username":"ellen_key", "password":"education"})

def index(l):
    l.client.get("/")

def profile(l):
    l.client.get("/profile")

class UserBehavior(TaskSet):
    tasks = {index:2, profile:1}

    def on_start(self):
        login(self)

class WebsiteUser(HttpLocust):
    task_set = UserBehavior
    min_wait = 5000
    max_wait = 10000

locust -f locustfile.py –host = http://192.168.1.4:8000

---

demo3:api test

#fil3.py
from locust import HttpLocust, TaskSet

headers= {
        'Cookie':'csrftoken=Ksrre3kIVCsQXv9g1II8Yt2grxj3C9KS; sessionid=aofv01n4v5piskrzm78dppj3erws2bn2'
        }

def api1(l):
    #print l.client.get('/api/v1/platform/strategy/live',headers=headers).json()
    print l.client.get('/api/v1/platform/strategy/live',headers=headers).json()['code']


def api2(l):
    print l.client.get('/api/v1/platform/strategy/account',headers=headers).json()['code']

class UserBehavior(TaskSet):
    tasks = {api1:2,api2:1}

    def on_start(self):
        print "begin"


class WebsiteUser(HttpLocust):
    task_set = UserBehavior
    min_wait=1000
    max_wait=2000

locust -f fil3.py –host:http://192.168.1.5

---

2.Apache Bench

ab

可以使用常用的Apache Bench.

$ab -n 100 -c 10 -l http://192.168.1.180/

The most common:

-n      requests Number of requests to perform
-c      concurrency  Number of multiple requests to make at a time
-l       Accept variable document length (use this for dynamic pages)

ab可以执行一些日常的负载测试，具体用法可参考ab的文档（man ab）。

siege

siege能比ab更真实的模拟实际的用户访问，且可以同时进行多个网页/网站的负载测试。

使用sproxy记录访问过程

为了模拟真实的用户访问过程，可以使用sproxy记录用户访问过程:
- 下载(sproxy)[http://download.joedog.org]作为代理服务器，
- 在浏览器上将代理服务器设置为sproxy。
- 模拟一个用户典型的访问流程。
- 从代理服务器sproxy上获得整个访问过程的URL记录。

$ sproxy -h
sproxy v1.02-March-20-2017
Usage:      sproxy [options] [hostname]
            (If a hostname is not specified, then sproxy will bind
            to localhost [127.0.0.1])
Options:
  -V        VERSION, prints version number to screen.
  -v        Verbose, prints URLS to screen.
  -h        Help, prints this section.
  -t NUM    Timeout, set the sproxy connection timeout to NUMM (default 120)
  -p NUM    Port, specify the port on which sproxy will listen for 
            incoming connections (default 9001)
  -f FILE   File, specify an alternative configuration file. 
            (default $prefix/etc/sproxy.conf)
  -o FILE   Output file, specify an alternative file to write URLs.
            (default $HOME/urls.txt)

$sproxy -p 9001 -o ./1.url

收集到的ruls访问记录包括：

$cat ./1.url
http://192.168.1.180/js/vendor/vue.min.js
http://192.168.1.180/js/vendor/echarts.min.js
http://192.168.1.180/js/my.bundle.js
http://192.168.1.180/js/strategy-upload.bundle.js
http://192.168.1.180/api/v1/user
http://192.168.1.180/api/v1/strategy_upload/strategy
http://192.168.1.180/api/v1/public/product
http://192.168.1.180:80
http://192.168.1.180/api/v1/strategy_upload/dependency
http://192.168.1.180/api/v1/strategy_upload/strategy POST ------WebKitFormBoundarywEDbBF...
http://192.168.1.180/api/v1/strategy_upload/strategy/100
http://192.168.1.180/api/v1/strategy_upload/strategy POST ------WebKitFormBoundaryMXmAkp...

为了测试真实的WEB服务进程压力，可以将sproxy收集到的url中的静态文件(nginx配置中静态文件)的urls去掉，
因为这部分的访问压力由Nginx承担，没有转发到后台的WEB服务进程。

配置siege进行测试

1.下载(siege)[http://download.joedog.org]
2. 使用记录的url列表进行测试：

$siege -f ./1.url -c 100 -r 10

可模拟GET/POST等请求

其它：

* 登陆和授权处理可参考：https://www.joedog.org/2012/02/16/http-authentication/

wrk

(wrk)[https://github.com/wg/wrk] 相比ab和siege是一个比较新的HTTP测试工具。

$wrk -t12 -c400 -d30s http://127.0.0.1:8080/index.html

其它

• https://github.com/yandex/yandex-tank
• https://github.com/naver/ngrinder
• http://jmeter.apache.org

参考：

1. en.wikipedia.org/wiki/Load_testing
2. https://www.joedog.org/sproxy-manual/