基於嵌入式webserver的伺服器狀態監控

其實也是在easyhadoop做第二次重構的時候用到了這個嵌入式的webserver去做伺服器狀態的監控，可以單獨摘出來寫個東西。

思路主要是用python指令碼擷取linux伺服器的各種狀態資訊，然後用webserver的方式，以json資料發給http，主控節點去讀取webserver返回的json，產生系統監控報表。代碼簡單，開發和部署都很方便。

用到的主要東西是python的第三方嵌入式web模組叫cherrypy，之所以選擇cherrypy，主要原因就是開發快速，學習也很快，基本我用了一天多就基本學會了怎麼來寫了。當然，也可以用python內建的simpleHTTPserver。不過那個確實太simple了。cherrypy的優點在於，多線程，多並發。又不像Tornado和Django那樣重量級。因為我們返回的是json，也用不到什麼html模板，資料庫的功能。當然也可以選擇web.py，不過相比還是cherrypy更好一點我認為，web.py也借鑒了cherrypy不少的思想。

其實我的叢集裡都裝了Ganglia, Cacti, Nagios。但是我想看即時的圖表產生，就自己寫了這樣一段程式。做自動化營運的朋友可以當一個參考，添加自己的方法，監控別的也是可以的。

主要流程是這樣：

想辦法讀取Linux系統的各種資料  --->  解析資料轉成json發給http伺服器  --->  監控伺服器掃描資料組建圖表

#!/usr/bin/python# -*- coding: utf8 -*-import sysimport cherrypyimport platformimport osimport time#python 2.4為simplejson，python 2.6以上為jsontry:    import jsonexcept ImportError:    import simplejson as json                                                                               #假裝做一個index出來class Index(object):    #下面這句表示修飾index方法，這個index方法是暴露給http server的    @cherrypy.expose    def index(self):        return "hello cherrypy"                                                                                   class Node(object):    '''    url /node/dist/    '''    #擷取目標機器的發行分支，版本號碼，架構類型，主機名稱等等，返回json    @cherrypy.expose    def 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/    '''    #擷取CPU型號等，返回json    @cherrypy.expose    def 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: continue            name = line.split(':')[0].strip().replace(' ','_')            var = line.split(':')[1].strip()            cpuinfo[name] = var        return json.dumps(cpuinfo, sort_keys=False, indent=4, separators=(',', ': '))    '''    url /node/GetMemInfo/    '''    #擷取記憶體使用量的詳細資料    @cherrypy.expose    def GetMemInfo(self):        mem = {}        f = open("/proc/meminfo")        lines = f.readlines()        f.close()        for line in lines:            if len(line) < 2:                   continue            name = line.split(':')[0]            var = line.split(':')[1].split()[0]            mem[name] = long(var) * 1024.0        mem['MemUsed'] = mem['MemTotal'] - mem['MemFree'] - mem['Buffers'] - mem['Cached']        return json.dumps(mem, sort_keys=False, indent=4, separators=(',', ': '))    '''    url /node/GetLoadAvg//    '''    #擷取系統負載的詳細資料    @cherrypy.expose    def 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)    '''    擷取指定網卡的流量資訊，這裡面有點複雜    @cherrypy.expose    def 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部分是給centos使用的，centos在流量大的情況下，網卡資訊裡面字串會連上，所以需要單獨拆分處理，else部分則是ubuntu或者其他系統格式化很好的使用            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)    #擷取全部網卡的介面和流量資訊    @cherrypy.expose    def 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:                continue        for interface in ifs:            nettraffic[interface] = self.GetIfInfo( interface)                                                                                           return json.dumps(nettraffic)    #擷取硬碟的分區資訊和使用量    @cherrypy.expose    def 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_system        mount['type'] = type        mount['size'] = size        mount['used'] = used        mount['avail'] = avail        mount['used_percent'] = used_percent        mount['mounted_on'] = mounted_on        dist_json = json.dumps(mount)        return dist_json                                                                                   #擷取CPU的使用量資訊，需要系統安裝sysstat支援    @cherrypy.expose    def 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 only                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\\\":\\\"\"$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 cpuif "__main__" == __name__:    #伺服器配置    settings = {                'global': {                    #綁定連接埠                    'server.socket_port' : 60090,                    #ip地址設定，覺得夠安全就用0.0.0.0，否則就單獨寫那台伺服器的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並啟動webserver        cherrypy.config.update(settings)        cherrypy.tree.mount(Index(), '/')        cherrypy.tree.mount(Node(), '/node')        cherrypy.engine.start()

圖表產生端隨便拿什麼語言寫就無所謂了，反正資料都是json格式的。

當然，我也用他監控hadoop和hbase。代碼加點跟hadoop和hbase相關的就可以了。

本文出自 “實踐檢驗真理” 部落格，謝絕轉載！