开源的运维系统不少，比如nagios、zabbix、cati等等，但是遇到自己个性化的运维需求的时候，总是显的力不从心！最近在学习python，所以就考虑用python+django+twisted来定做一个完全个性化的运维系统。

运维系统有几个主要的功能：监控、分析、报警、更甚者直接根据分析的结果进行反应操作。而以上几点通过上述的框架可以比较容易的实现。

下面上图说明：

使用freemind整理了下思路：

下面是一些代码段，完整的代码下载见文档底部：

Server：

#!/usr/bin/env python
#coding:utf-8
__author__ = 'dwj'from twisted.internet.protocol import ServerFactory
from twisted.protocols import basic
import cx_Oracle
from twisted.application import  service, internetclass Mornitor_Protocol(basic.LineReceiver):def __init__(self):#_oracle_conn=cx_Oracle.connect('xxxx', 'xxxx', '192.168.7.17/test', threaded=True)_oracle_conn.autocommit = Trueself.cur = _oracle_conn.cursor()self._oracle_conn=_oracle_conndef ruku(self, line):ip=self.transport.getPeer().host#获取客户端IPline=line.split(':::')#使用：：：分割原始数据if line[1] in ['cpu', 'mem', 'disk', 'tcp', 'net', 'process_down']:#根据数据包头来确定使用insert还是update，当是tcp包头的时候插入，其余的更新if line[1] == 'tcp':sql = "insert into MORNITOR_BASICINFO (ipadd,time,tcp) values (\'%s\',\'%s\',\'%s\')"%(ip,line[0],line[3])print sqlself.cur.execute(sql)else:line_again = line[3].split('::')sql = 'update MORNITOR_BASICINFO set %s=\'%s\',%s=\'%s\' where ipadd=\'%s\' and time=\'%s\''%(line[1],line_again[0],line[2],line_again[1],ip,line[0])print sqlself.cur.execute(sql)def connectionMade(self):print 'Connected!'def lineReceived(self, line):print lineself.ruku(line)#接受到数据之后执行入库操作！def connectionLost(self, reason='connectionDone'):self._oracle_conn.close()print 'The db is close... ok!'class Mornitor_Factory(ServerFactory):#还没想好要初始化什么def __init__(self,service):self.service = serviceprotocol = Mornitor_Protocolclass Fish_Service(service.Service):def __init__(self):passdef startService(self):service.Service.startService(self)              #什么都不做，开始服务# def stopService(self):#     return self._port.stopListening()#配置参数
port = 10000
iface = '127.0.0.1'top_server = service.MultiService()                             #定义服务容器fish_server = Fish_Service()                                    #实例化我们的服务
fish_server.setServiceParent(top_server)                        #把自定义的服务加入到服务容器factory = Mornitor_Factory(Fish_Service)                        #工厂化服务tcp_server = internet.TCPServer(port, factory, interface=iface) #定义tcp服务
tcp_server.setServiceParent(top_server)                         #把tcp服务加入到服务容器application = service.Application('Fish_Service')               #给应用起个名字
top_server.setServiceParent(application)                        #把服务容器丢到应用中去

Client端

from twisted.protocols import basic
from twisted.internet import  protocol, defer, task
import Get_basic_info_2 as Huoqu
import guardian as shouhu
import time
from twisted.application import service, internetclass Monitor_Protocol(basic.LineReceiver):#自定义客户端和服务端的连接协议，从basic的line继承def __init__(self):#pass@staticmethoddef huoqu_shuju():#定义一个函数获取本机的一些状态now = str(time.strftime('%Y-%m-%d %H:%M:%S'))def add_tag(source, tag1, tag2 = 'none'):#定义格式化字符串函数return ':::'.join([now, tag1, tag2, source])#使用：：：分隔时间、简单信息、详细信息、原始信息tcp = add_tag(Huoqu.net_tcp(), 'tcp')cpu = add_tag(Huoqu.cpu(), 'cpu', 'cpu_detail')mem = add_tag(Huoqu.mem(), 'mem', 'mem_detail')disk = add_tag(Huoqu.disk_usage(), 'disk', 'disk_detail')net = add_tag(Huoqu.net_rate(), 'net', 'net_detail')process = add_tag(shouhu.check_alive(), 'process_down', 'process_alived')result = (tcp, cpu, mem, disk, net, process, ) d = defer.Deferred()#使用defered返回结果d.callback(result)return ddef xunhuan(self, list):#定义循环发送函数for i in list:self.sendLine(i)def fasong(self):#定义程序运行顺序，取得信息后用callback交给发送函数发送self.huoqu_shuju().addCallback(self.xunhuan)def loop(self):#使用twist内置的循环函数定义几秒监控数据传送到服务端l = task.LoopingCall(self.fasong)l.start(1)def connectionMade(self):#覆盖协议的connectmade函数，定义于服务端的连接建立后开始循环print 'Connected!......ok!'self.loop()def lineReceived(self, line):#必须覆盖接受函数，否则twist会报not importent错误！passclass Moinitor_client_factory(protocol.ReconnectingClientFactory):def __init__(self, service):#还没想要要写什么self.service = serviceprotocol = Monitor_Protocolclass Client_Service(service.Service):def __init__(self):passdef startService(self):service.Service.startService(self)#配置文件开始
port = 10000
host = '127.0.0.1'#守护进程
top_service = service.MultiService()                   #定义服务容器client_service = Client_Service()                      #实例化服务类
client_service.setServiceParent(top_service)           #把自己定义的服务丢到服务容器中factory = Moinitor_client_factory(client_service)      #定义服务工厂化tcp_service = internet.TCPClient(host, port, factory)  #定义tcp连接的服务
tcp_service.setServiceParent(top_service)              #把tcp服务丢到服务容器中去application = service.Application('Fish_Service')      #定义应用名字
top_service.setServiceParent(application)              #把服务容器丢到应用中去

一些自定义监控程序是否存活的脚本：

program = {'nginx': ['/opt/nginx/logs/nginx.pid', '/opt/nginx/sbin/nginx'],'rsync-C': ['/var/run/rsyncd.pid', 'rsync --daemon'],}def main():for k in program:a = get_pid(k, program[k][0])if isinstance(a, tuple):print '%s is not running!' % kprint 'Start the program by Horland_guardian!'subprocess.call(program[k][1], shell=True)else:print 'The %s is running!' % kdef check_alive():l_lived = []l_downed = []for k in program:a = get_pid(k, program[k][0])if isinstance(a, tuple):l_downed.append(k)else:l_lived.append(k)process_alived = ' '.join(l_lived)process_down = ' '.join(l_downed)return '::'.join([process_down, process_alived])

django的使用目前只需要使用到admin模块就可以。

下面是一些代码段：

model

class BasicInfo(models.Model):ipadd = models.IPAddressField(verbose_name = u'IP地址')time = models.CharField(max_length=50, verbose_name = u'时间')cpu = models.CharField(max_length=255, blank=True, verbose_name = u'CPU%')cpu_detail = models.CharField(max_length=255, blank=True, verbose_name = u'CPU详情')mem = models.CharField(max_length=255, blank=True, verbose_name = u'内存%')mem_detail = models.CharField(max_length=255, blank=True, verbose_name = u'内存详情')disk = models.CharField(max_length=255, blank=True, verbose_name = u'磁盘%')disk_detail = models.CharField(max_length=255, blank=True, verbose_name = u'磁盘详情')net = models.CharField(max_length=255, blank=True, verbose_name = u'流量 bytes/s')net_detail = models.CharField(max_length=1000, blank=True, verbose_name = u'流量详情')tcp = models.CharField(max_length=255, blank=True, verbose_name = u'tcp连接状态')process_down = models.CharField(max_length=255, blank=True, verbose_name = u'DOWN-进程')process_alived = models.CharField(max_length=255, blank=True, verbose_name = u'Process_UP')def Process_DOWN(self):return '<span style="color: #%s;">%s</span>' % ('ff0000', self.process_down)  #拓机的进程用红色标识Process_DOWN.allow_tags = True

注册到admin

class BasicInfo_admin(admin.ModelAdmin):list_display = ('time', 'cpu', 'cpu_detail', 'mem', 'mem_detail', 'disk', 'disk_detail', 'net', 'net_detail', 'tcp', 'Process_DOWN', 'process_alived')list_filter = ('ipadd', )
admin.site.register(BasicInfo, BasicInfo_admin)

freemind整理的思路中还有一些功能没有实现，目前这个只能算个简单的demon吧，但是基本实现了监控的目的。欢迎大家给我留言！

下面上个django的admin界面截图吧！

代码下载

http://download.csdn.net/detail/qcpm1983/7611579