Python性能监控Graphite

rgfsd2

一、简介
Graphite 是一个Python写的web应用，采用django框架，Graphite用来进行收集服务器所有的及时状态，用户请求信息，Memcached命中率，RabbitMQ消息服务器的状态，Unix操作系统的负载状态，Graphite服务器大约每分钟需要有4800次更新操作，Graphite采用简单的文本协议和绘图功能可以方便地使用在任何操作系统上。

graphite有三个组件：

• graphite-web：web接口
  
• carbon：相当于network interface
  
• whisper：相当于rrdtool
  
  

graphite官方文档：
http://graphite.wikidot.com/documentation

http://graphite.readthedocs.org/en/latest/

二、安装graphite
1、安装epel源

1 2 3

rpm -ivh http://dl.fedoraproject.org/pub/ ... ease-6-8.noarch.rpm sed -i 's@^#@@' /etc/yum.repos.d/epel.repo sed -i 's@mirrorlist@#mirrorlist@' /etc/yum.repos.d/epel.repo

2、安装适应版本的Django软件包，版本过高会出现bug

1 2 3

yum install python-simplejson wget https://kojipkgs.fedoraproject.o ... 14-1.el6.noarch.rpm rpm -ivh Django14-1.4.14-1.el6.noarch.rpm

3、安装graphite

1	yum install graphite-web python-carbon python-whisper

4、安装MySQL数据库

1 2 3 4 5

yum install mysql mysql-server MySQL-python service mysqld start chkconfig mysqld on mysqladmin -uroot password 123456 mysql -uroot -p123456 -e 'create database graphite;'

5、修改graphite配置文件

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

# cat >> /etc/graphite-web/local_settings.py << EOF SECRET_KEY = '123qwe' ALLOWED_HOSTS = [ '*' ] TIME_ZONE = 'Asia/Shanghai' DEBUG = True DATABASES = {     'default': {         'NAME': 'graphite',         'ENGINE': 'django.db.backends.mysql',         'USER': 'root',         'PASSWORD': '123456',         'HOST': '127.0.0.1',         'PORT': '3306'     } } from graphite.app_settings import * EOF

6、同步数据库

1 2 3 4 5 6 7 8

mkdir -p /opt/graphite/storage cd /etc/graphite-web/ django-admin syncdb --settings=local_settings --pythonpath=. yes root zhengys@allentuns.com 123456 123456

7、修改graphite数据目录

1	chown -R apache.apache /opt/graphite/storage

8、启动服务

1 2 3 4

/etc/init.d/carbon-cache start chkconfig carbon-cache on /etc/init.d/httpd start chkconfig httpd on

三、访问展示graphite
1、Chrome浏览器访问Ghipte的地址：

2、提供监控网卡流量的脚本

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

[iyunv@Allentuns ~]# cat network_traffic.py #!/usr/bin/env python from subprocess import Popen,PIPE import socket import shlex import time import sys import os def get_traffic(f):     p = Popen(shlex.split(f),stdout=PIPE,stderr=PIPE)     result = p.stdout.read()     d = [i for i in  result.split('\n')[3:] if i]     dic_traffic = {}     for i in d:         devname = i.split(':')[0].strip()         Receive = i.split(':')[1].split()[0].strip()         Transmit = i.split(':')[1].split()[8].strip()         dic_traffic[devname] = {'in':Receive,'out':Transmit}     return dic_traffic if __name__ == '__main__':     try:         HOST = '127.0.0.1'     PORT = 2003         s = socket.socket()         s.connect((HOST,PORT))     except:     print "Couldn't connect to %(server)s on port %(port)d, is carbon-agent.py running?" % {'server':HOST,'port':PORT}     sys.exit(1)     while True:         cur_traffic = get_traffic('cat /proc/net/dev')         time.sleep(5)         five_s_traffic = get_traffic('cat /proc/net/dev')         diff_dic = {}         for k in cur_traffic:         traffic_in = int(five_s_traffic[k]['in']) - int(cur_traffic[k]['in'])         traffic_out = int(five_s_traffic[k]['out']) - int(cur_traffic[k]['out'])         diff_dic[k] = {'in':traffic_in,'out':traffic_out}     now = int(time.time())     for k,v in diff_dic.items():         net_in = 'network.%s_in %s %s\n' % (k,v['in'],now)         net_out = 'network.%s_out %s %s\n' % (k,v['out'],now)         s.sendall(net_in)         s.sendall(net_out)     time.sleep(5)

3、后台方式运行监控网卡流量脚本

1	[iyunv@Allentuns ~]# python network_traffic.py &

四、安装Diamond

diamond ：搜集器、用于搜集数据
diamond的github官方站点：https://github.com/python-diamond/Diamond/wiki
1、安装Diamond

1 2 3 4 5 6 7

yum install gcc gcc-c++ python-configobj python-pip pip install diamond==3.4.421  (有时候会安装不成功) 如果下载安装不成功可以使用以下方式进行 wget https://pypi.python.org/packages ... d81f16a4fd27d11093a tar xf diamond-3.4.421.tar.gz cd diamond-3.4.421 python setup.py install

2、配置

1 2 3 4 5 6 7 8 9 10 11 12 13

cd /etc/diamond/ cp diamond.conf.example diamond.conf 主要修改三个配置文件： [iyunv@Allentuns diamond]# vim /etc/diamond/diamond.conf [[GraphiteHandler]]    //59行 host = localhost [[default]]         //173行 interval = 10     //时间搜集一次 [iyunv@Allentuns diamond]# vim /etc/diamond/handlers/ArchiveHandler.conf #log_file = ./storage/archive.log //注释此行 [iyunv@Allentuns diamond]# vim /etc/diamond/handlers/GraphiteHandler.conf host = localhost

3、启动diamond服务

1 2 3

chmod +x /etc/init.d/diamond /etc/init.d/diamond start chkconfig diamond on

五、继续访问展示diamond自动采集信息
1、Chrome浏览器访问Ghipte的地址：
你会发现在Graphite下多了一个servers的目录，这个目录就是diamond自动采集的信息
2、在这里提供了两个python脚本，用来搜集web站点的httpcode，是基于diamond的方式

1 2 3 4

[iyunv@Allentuns httpcode]# ll 总用量 8 -rwxr-xr-x 1 root root 1356 3月  31 11:12 filerev.py -rwxr-xr-x 1 root root 3737 3月  31 11:12 httpcode.py

3、运行搜集httpcode的脚本

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

首先删除原来diamond生成的servers目录 [iyunv@Allentuns httpcode]# rm -rf /var/lib/carbon/whisper/servers/ 然后手动运行diamond的httpcode脚本 [iyunv@Allentuns httpcode]# diamond -f -l -r ./httpcode.py  -c /etc/diamond/diamond.conf ERROR: Pidfile exists. Server already running?     #需要手动停止diamond服务 [iyunv@Allentuns httpcode]# /etc/init.d/diamond stop Stopping diamond:                                          [确定] [iyunv@Allentuns httpcode]# diamond -f -l -r ./httpcode.py  -c /etc/diamond/diamond.conf [2015-03-31 11:13:56,198] [MainThread] Changed UID: 0 () GID: 0 (). [2015-03-31 11:13:56,198] [MainThread] Loaded Handler: diamond.handler.graphite.GraphiteHandler [2015-03-31 11:13:56,201] [MainThread] GraphiteHandler: Established connection to graphite server localhost:2003. [2015-03-31 11:13:56,202] [MainThread] Loaded Handler: diamond.handler.archive.ArchiveHandler [2015-03-31 11:13:56,206] [MainThread] Loading Collectors from: . [2015-03-31 11:13:56,209] [MainThread] Loaded Module: httpcode [2015-03-31 11:13:56,209] [MainThread] Loaded Collector: httpcode.HttpCodeCollector [2015-03-31 11:13:56,209] [MainThread] Initialized Collector: HttpCodeCollector [2015-03-31 11:13:56,210] [MainThread] Skipped loading disabled Collector: HttpCodeCollector [2015-03-31 11:13:56,210] [MainThread] Started task scheduler. [2015-03-31 11:13:57,211] [MainThread] Stopping task scheduler. [2015-03-31 11:14:01,217] [MainThread] Stopped task scheduler. [2015-03-31 11:14:01,217] [MainThread] Exiting. 如果没有报错，则查看浏览器会发现多了一个servers目录；但是当时目录就是没有生成，我还真纳闷了。原来在配置文件中没有启动此配置 [iyunv@Allentuns httpcode]# vim httpcode.py ...... config = super(HttpCodeCollector, self).get_default_config()         config.update({             'path':     'weblog',             'enabled':  'True'    #开启此选项         }) 如果用diamond来搜集，则无需此选项，因为diamond有针对类的配置文件，在配置文件中开启会比在脚本中开启看起来更统一

4、浏览器查看
Chrome刷新Ghipte的web页面，查看
Ghipte -> servers -> ec2-54-201-82-69 -> weblog(自定义) -> http 会出现以下监控曲线图
我们可以使用ab -c 100 -n 100 http://localhost/ 产生200状态码
使用刷新Ghipte的浏览器页面产生304的状态码

• 目前主流的开源监控有Cacti、Nagios、Zabbix等等，社区活跃，功能强大
  
• Graphite虽然在功能上和社区在无法与此对比，但是在灵活度上还是值得一提的，轻量级的监控程序，更为重要的是Graphite是Python编写的，所以在问题排查，脚本编写等都会非常顺手
  
• 同样也非常感谢更多Python开源者的贡献！！！