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
rpm -ivh http://dl.fedoraproject.org/pub/epel/6/x86_64/epel-release-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
yum install python-simplejson wget https://kojipkgs.fedoraproject.org//packages/Django14/1.4.14/1.el6/noarch/Django14-1.4.14-1.el6.noarch.rpm rpm -ivh Django14-1.4.14-1.el6.noarch.rpm
3、安装graphite
yum install graphite-web python-carbon python-whisper
4、安装MySQL数据库
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
# 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
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
chown -R apache.apache /opt/graphite/storage
8、启动服务
/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
[root@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
[root@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
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/source/d/diamond/diamond-3.4.421.tar.gz#md5=080ab9f52a154d81f16a4fd27d11093a tar xf diamond-3.4.421.tar.gz cd diamond-3.4.421 python setup.py install
2、配置
cd /etc/diamond/ cp diamond.conf.example diamond.conf 主要修改三个配置文件: [root@Allentuns diamond]# vim /etc/diamond/diamond.conf [[GraphiteHandler]] //59行 host = localhost [[default]] //173行 interval = 10 //时间搜集一次 [root@Allentuns diamond]# vim /etc/diamond/handlers/ArchiveHandler.conf #log_file = ./storage/archive.log //注释此行 [root@Allentuns diamond]# vim /etc/diamond/handlers/GraphiteHandler.conf host = localhost
3、启动diamond服务
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
[root@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
首先删除原来diamond生成的servers目录 [root@Allentuns httpcode]# rm -rf /var/lib/carbon/whisper/servers/ 然后手动运行diamond的httpcode脚本 [root@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 [root@Allentuns httpcode]# /etc/init.d/diamond stop Stopping diamond: [确定] [root@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 [root@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!!!
本文出自 “郑彦生” 博客,请务必保留此出处http://467754239.blog.51cto.com/4878013/1626736
原文地址:http://467754239.blog.51cto.com/4878013/1626736