其实也是在easyhadoop做第二次重构的时候用到了这个嵌入式的webserver去做服务器状态的监控，可以单独摘出来写个东西。

思路主要是用python脚本获取linux服务器的各种状态信息，然后用webserver的方式，以json数据发给http，主控节点去读取webserver返回的json，生成系统监控报表。代码简单，开发和部署都很方便。

用到的主要东西是python的第三方嵌入式web模块叫cherrypy，之所以选择cherrypy，主要原因就是开发快速，学习也很快，基本我用了一天多就基本学会了怎么来写了。当然，也可以用python自带的simpleHTTPserver。不过那个确实太simple了。cherrypy的优点在于，多线程，多并发。又不像Tornado和Django那样重量级。因为我们返回的是json，也用不到什么html模板，数据库的功能。当然也可以选择web.py，不过相比还是cherrypy更好一点我认为，web.py也借鉴了cherrypy不少的思想。

其实我的集群里都装了Ganglia, Cacti, Nagios。但是我想看实时的图表生成，就自己写了这样一段程序。做自动化运维的朋友可以当一个参考，添加自己的方法，监控别的也是可以的。

主要流程是这样：

想办法读取Linux系统的各种数据  --->  解析数据转成json发给http服务器  --->  监控服务器扫描数据生成图表

#!/usr/bin/python
# -*- coding: utf8 -*-
import sys
import cherrypy
import platform
import os
import time
#python 2.4为simplejson，python 2.6以上为json
try:import json
except ImportError:import simplejson as json#假装做一个index出来
class Index(object):#下面这句表示修饰index方法，这个index方法是暴露给http server的@cherrypy.exposedef index(self):return "hello cherrypy"class Node(object):'''url /node/dist/'''#获取目标机器的发行分支，版本号，架构类型，主机名称等等，返回json@cherrypy.exposedef dist(self):dist_json = ''sysinstaller = ''installer = ''ostype = platform.dist()if(ostype[0] in ['Ubuntu','debian','ubuntu','Debian']):sysinstaller = 'apt-get'installer = 'dpkg'elif(ostype[0] in ['SuSE']):sysinstaller = 'zypper'installer = 'rpm'elif(ostype[0] in ['CentOS', 'centos', 'redhat','RedHat']):sysinstaller = 'yum'installer = 'rpm'machine = platform.machine()hostname = platform.node()dist_json = {'os.system':ostype[0], 'os.version':ostype[1], 'os.release':ostype[2], 'os.sysinstall':sysinstaller, 'os.installer':installer, 'os.arch':machine, 'os.hostname':hostname}return json.dumps(dist_json, sort_keys=False, indent=4, separators=(',', ': '))'''url /node/GetCpuInfo/'''#获取CPU型号等，返回json@cherrypy.exposedef GetCpuInfo(self):cpu = []cpuinfo = {}f = open("/proc/cpuinfo")lines = f.readlines()f.close()for line in lines:if line == 'n':cpu.append(cpuinfo)cpuinfo = {}if len(line) < 2: continuename = line.split(':')[0].strip().replace(' ','_')var = line.split(':')[1].strip()cpuinfo[name] = varreturn json.dumps(cpuinfo, sort_keys=False, indent=4, separators=(',', ': '))'''url /node/GetMemInfo/'''#获取内存使用的详细信息@cherrypy.exposedef GetMemInfo(self):mem = {}f = open("/proc/meminfo")lines = f.readlines()f.close()for line in lines:if len(line) < 2:continuename = line.split(':')[0]var = line.split(':')[1].split()[0]mem[name] = long(var) * 1024.0mem['MemUsed'] = mem['MemTotal'] - mem['MemFree'] - mem['Buffers'] - mem['Cached']return json.dumps(mem, sort_keys=False, indent=4, separators=(',', ': '))'''url /node/GetLoadAvg//'''#获取系统负载的详细信息@cherrypy.exposedef GetLoadAvg(self):loadavg = {}f = open("/proc/loadavg")con = f.read().split()f.close()loadavg['lavg_1']=con[0]loadavg['lavg_5']=con[1]loadavg['lavg_15']=con[2]loadavg['nr']=con[3]loadavg['last_pid']=con[4]return json.dumps(loadavg, sort_keys=False, indent=4, separators=(',', ': '))'''url /node/GetIfInfo/eth(x)'''获取指定网卡的流量信息，这里面有点复杂@cherrypy.exposedef GetIfInfo(self, interface):dist_json = self.dist()f = open("/proc/net/dev")lines = f.readlines()f.close()intf = {}for line in lines[2:]:con = line.split()#if部分是给centos使用的，centos在流量大的情况下，网卡信息里面字符串会连上，所以需要单独拆分处理，else部分则是ubuntu或者其他系统格式化很好的使用if con[0][-1].isdigit() == True:offset = con[0].split(':')intf['interface'] = str(offset[0])intf['ReceiveBytes'] = str(offset[1])intf['ReceivePackets'] = str(con[1])intf['ReceiveErrs'] = str(con[2])intf['ReceiveDrop'] = str(con[3])intf['ReceiveFifo'] = str(con[4])intf['ReceiveFrames'] = str(con[5])intf['ReceiveCompressed'] = str(con[6])intf['ReceiveMulticast'] = str(con[7])intf['TransmitBytes'] = str(con[8])intf['TransmitPackets'] = str(con[9])intf['TransmitErrs'] = str(con[10])intf['TransmitDrop'] = str(con[11])intf['TransmitFifo'] = str(con[12])intf['TransmitFrames'] = str(con[13])intf['TransmitCompressed'] = str(con[14])intf['TransmitMulticast'] = str(con[15])else:intf['interface'] = str(con[0])intf['ReceiveBytes'] = str(con[1])intf['ReceivePackets'] = str(con[2])intf['ReceiveErrs'] = str(con[3])intf['ReceiveDrop'] = str(con[4])intf['ReceiveFifo'] = str(con[5])intf['ReceiveFrames'] = str(con[6])intf['ReceiveCompressed'] = str(con[7])intf['ReceiveMulticast'] = str(con[8])intf['TransmitBytes'] = str(con[9])intf['TransmitPackets'] = str(con[10])intf['TransmitErrs'] = str(con[11])intf['TransmitDrop'] = str(con[12])intf['TransmitFifo'] = str(con[13])intf['TransmitFrames'] = str(con[14])intf['TransmitCompressed'] = str(con[15])intf['TransmitMulticast'] = str(con[16])return json.dumps(intf, sort_keys=False)#获取全部网卡的接口和流量信息@cherrypy.exposedef GetIfTraffic(self):ifs = []nettraffic = {}f = open("/proc/net/dev")lines = f.readlines()f.close()for line in lines[2:]:con = line.split()ifname = con[0].split(':')if(ifname[0].strip() != 'lo'):ifs.append(ifname[0].strip())else:continuefor interface in ifs:nettraffic[interface] = self.GetIfInfo( interface)return json.dumps(nettraffic)#获取硬盘的分区信息和使用量@cherrypy.exposedef GetHddInfo(self):hdds = []mount = {}file_system = []type = []size = []used = []avail = []used_percent = []mounted_on = []hdds = os.popen('df -lhT  | grep -v tmpfs | grep -v boot | grep -v usr | grep -v tmp | sed \'1d;/ /!N;s/\\n//;s/[ ]*[ ]/\\t/g;\'').readlines()for line in hdds:file_system.append(line.replace('\\n','').replace('\\t',' ').split()[0])type.append(line.replace('\\n','').replace('\\t',' ').split()[1])size.append(line.replace('\\n','').replace('\\t',' ').split()[2])used.append(line.replace('\\n','').replace('\\t',' ').split()[3])avail.append(line.replace('\\n','').replace('\\t',' ').split()[4])used_percent.append(line.replace('\\n','').replace('\\t',' ').split()[5])mounted_on.append(line.replace('\\n','').replace('\\t',' ').split()[6])mount['file_system'] = file_systemmount['type'] = typemount['size'] = sizemount['used'] = usedmount['avail'] = availmount['used_percent'] = used_percentmount['mounted_on'] = mounted_ondist_json = json.dumps(mount)return dist_json#获取CPU的使用量信息，需要系统安装sysstat支持@cherrypy.exposedef GetCpuDetail(self):dist_json = self.dist()dist = json.loads(dist_json)if(dist['os.system'] in ['CentOS', 'centos', 'redhat', 'RedHat']):if(int(dist['os.version'].split('.')[0])  < 6):  #For CentOS onlycmd = 'mpstat 1 1 | sed \'1d;2d;3d;4d\' | awk \'{print "{\\\"user\\\":\\\"\"$3\"\\\",\\\"nice\\\":\\\"\"$4\"\\\",\\\"sys\\\":\\\"\"$5\"\\\",\\\"iowait\\\":\\\"\"$6\"\\\",\\\"irq\\\":\\\"\"$7\"\\\",\\\"soft\\\":\\\"\"$8\"\\\",\\\"steal\\\":\\\"\"$9\"\\\",\\\"idle\\\":\\\"\"$10\"\\\"}"}\''else:cmd = 'mpstat 1 1 | sed \'1d;2d;3d;4d\' | awk \'{print "{\\\"user\\\":\\\"\"$3\"\\\",\\\"nice\\\":\\\"\"$4\"\\\",\\\"sys\\\":\\\"\"$5\"\\\",\\\"iowait\\\":\\\"\"$6\"\\\",\\\"irq\\\":\\\"\"$7\"\\\",\\\"soft\\\":\\\"\"$8\"\\\",\\\"steal\\\":\\\"\"$9\"\\\",\\\"idle\\\":\\\"\"$11\"\\\"}"}\''else:cmd = 'mpstat 1 1 | sed \'1d;2d;3d;4d\' | awk \'{print "{\\\"user\\\":\\\"\"$3\"\\\",\\\"nice\\\":\\\"\"$4\"\\\",\\\"sys\\\":\\\"\"$5\"\\\",\\\"iowait\\\":\\\"\"$6\"\\\",\\\"irq\\\":\\\"\"$7\"\\\",\\\"soft\\\":\\\"\"$8\"\\\",\\\"steal\\\":\\\"\"$9\"\\\",\\\"idle\\\":\\\"\"$11\"\\\"}"}\''cpu = os.popen(cmd).readline().strip()return cpu
if "__main__" == __name__:#服务器配置settings = {'global': {#绑定端口'server.socket_port' : 60090,#ip地址设置，觉得够安全就用0.0.0.0，否则就单独写那台服务器的ip'server.socket_host': '0.0.0.0','server.socket_file': '','server.socket_queue_size': 100,'server.protocol_version': 'HTTP/1.1','server.log_to_screen': True,'server.log_file': '','server.reverse_dns': False,'server.thread_pool': 200,'server.environment': 'production','engine.timeout_monitor.on': False}}#使用配置和映射路由并启动webservercherrypy.config.update(settings)cherrypy.tree.mount(Index(), '/')cherrypy.tree.mount(Node(), '/node')cherrypy.engine.start()

图表生成端随便拿什么语言写就无所谓了，反正数据都是json格式的。

当然，我也用他监控hadoop和hbase。代码加点跟hadoop和hbase相关的就可以了。