zabbix企业应用之如何快速查看异常流量
最近机房总流量总是异常，然后我是不断的接到短信与电话报警，收到后通过cacti查看还是十分的麻烦与浪费时间，为了解决这个问题，我自己写了一个脚本，从数据库里获取所有主机监控数据，然后打印出流量超过10m的主机信息，这样能快速帮我判断异常流量主机。
脚本是使用python编写，使用MySQLdb从zabbix数据库里获取流量数据，经过流量判断后在把数据写入到excel里。
使用前需要安装MySQLdb、xlwt模块，可以使用easy_install安装。
下面是脚本内容
#!/usr/bin/env python #-*- coding: utf-8 -*- #author:Deng Lei #email: [email protected] import MySQLdb import time import sys import xlwt reload(sys) sys.setdefaultencoding('utf8') if __name__ == "__main__": now_hour=int(time.strftime('%H')) old_hour=now_hour-1 now_time=time.strftime('%Y-%m-%d') mysql_conn=MySQLdb.connect(host='10.10.14.11',user='zabbix',passwd='zabbix',port=3306,charset="utf8") mysql_cur=mysql_conn.cursor() mysql_conn.select_db('zabbix') last_results=[] in_results=[] out_results=[] network_device=['em2','eth1','eth0'] try: room=sys.argv[1] except IndexError: room='all' try: old_hour=sys.argv[2] except IndexError: old_hour=now_hour-1 try: now_hour=sys.argv[3] except IndexError: now_hour=int(time.strftime('%H')) if room == 'all': #search network in traffic for i in network_device: search_sql="select from_unixtime(hi.clock,'%%Y-%%m-%%d %%T') as Date,g.name as Group_Name,h.host as Host,round(max(hi.value_max)/1000,0) as Network from hosts_groups hg join groups g on g.groupid = hg.groupid join items i on hg.hostid = i.hostid join hosts h on h.hostid=i.hostid join trends_uint hi on i.itemid = hi.itemid where i.key_='net.if.in[%s]' and hi.clock >= UNIX_TIMESTAMP('%s %s:00:00') and hi.clock < UNIX_TIMESTAMP('%s %s:00:00') group by h.host;"%(i,now_time,old_hour,now_time,now_hour) n=mysql_cur.execute(search_sql) result=mysql_cur.fetchall() for ii in result: msg1={'Group_Name':ii[1],'Host':ii[2],'Network_device':i,'Source':'In'} if msg1 not in in_results: in_results.append(msg1) msg={'Date':ii[0],'Group_Name':ii[1],'Host':ii[2],'Network':float(ii[3]),'Network_device':i,'Source':'In'} last_results.append(msg) #search network out traffic for i in network_device: search_sql="select from_unixtime(hi.clock,'%%Y-%%m-%%d %%T') as Date,g.name as Group_Name,h.host as Host,round(max(hi.value_max)/1000,0) as Network from hosts_groups hg join groups g on g.groupid = hg.groupid join items i on hg.hostid = i.hostid join hosts h on h.hostid=i.hostid join trends_uint hi on i.itemid = hi.itemid where i.key_='net.if.out[%s]' and hi.clock >= UNIX_TIMESTAMP('%s %s:00:00') and hi.clock < UNIX_TIMESTAMP('%s %s:00:00') group by h.host;"%(i,now_time,old_hour,now_time,now_hour) n=mysql_cur.execute(search_sql) result=mysql_cur.fetchall() for ii in result: msg1={'Group_Name':ii[1],'Host':ii[2],'Network_device':i,'Source':'out'} if msg1 not in out_results: out_results.append(msg1) msg={'Date':ii[0],'Group_Name':ii[1],'Host':ii[2],'Network':float(ii[3]),'Network_device':i,'Source':'out'} last_results.append(msg) else: #search network in traffic for i in network_device: search_sql="select from_unixtime(hi.clock,'%%Y-%%m-%%d %%T') as Date,g.name as Group_Name,h.host as Host,round(max(hi.value_max)/1000,0) as Network from hosts_groups hg join groups g on g.groupid = hg.groupid join items i on hg.hostid = i.hostid join hosts h on h.hostid=i.hostid join trends_uint hi on i.itemid = hi.itemid where i.key_='net.if.in[%s]' and g.name like '%s' and hi.clock >= UNIX_TIMESTAMP('%s %s:00:00') and hi.clock < UNIX_TIMESTAMP('%s %s:00:00') group by h.host;"%(i,room+"%",now_time,old_hour,now_time,now_hour) n=mysql_cur.execute(search_sql) result=mysql_cur.fetchall() for ii in result: msg1={'Group_Name':ii[1],'Host':ii[2],'Network_device':i,'Source':'In'} if msg1 not in in_results: in_results.append(msg1) msg={'Date':ii[0],'Group_Name':ii[1],'Host':ii[2],'Network':float(ii[3]),'Network_device':i,'Source':'In'} last_results.append(msg) #search network out traffic for i in network_device: search_sql="select from_unixtime(hi.clock,'%%Y-%%m-%%d %%T') as Date,g.name as Group_Name,h.host as Host,round(max(hi.value_max)/1000,0) as Network from hosts_groups hg join groups g on g.groupid = hg.groupid join items i on hg.hostid = i.hostid join hosts h on h.hostid=i.hostid join trends_uint hi on i.itemid = hi.itemid where i.key_='net.if.out[%s]' and g.name like '%s' and hi.clock >= UNIX_TIMESTAMP('%s %s:00:00') and hi.clock < UNIX_TIMESTAMP('%s %s:00:00') group by h.host;"%(i,room+"%",now_time,old_hour,now_time,now_hour) n=mysql_cur.execute(search_sql) result=mysql_cur.fetchall() for ii in result: msg1={'Group_Name':ii[1],'Host':ii[2],'Network_device':i,'Source':'out'} if msg1 not in out_results: out_results.append(msg1) msg={'Date':ii[0],'Group_Name':ii[1],'Host':ii[2],'Network':float(ii[3]),'Network_device':i,'Source':'out'} last_results.append(msg) time="%s-[%s-%s]"%(now_time,old_hour,now_hour) a=[] for i in last_results: if i['Network'] >=10000: msg=(i['Group_Name'],i['Host'],i['Network']/1000,i['Network_device'],i['Source']) a.append(msg) sort_list=sorted(a,key=lambda d:d[2],reverse = True) wb = xlwt.Workbook() ws = wb.add_sheet('zabbix', cell_overwrite_ok=True) ws.write(0,0,'报警组'.decode("utf-8")) ws.write(0,1,'主机'.decode("utf-8")) ws.write(0,2,'流量(Mbps)'.decode("utf-8")) ws.write(0,3,'网卡名'.decode("utf-8")) ws.write(0,4,'方向'.decode("utf-8")) for i in range(1,len(sort_list)+1): for ii in range(0,len(sort_list[i-1])): ws.write(i,ii,sort_list[i-1][ii]) ws.col(0).width = 3333*3 ws.col(1).width = 3333 wb.save('/tmp/zabbix_network_traffic-%s.xls'%time) mysql_cur.close() mysql_conn.close()
运行的话，参数信息如下：
第一个参数是机房信息，比如我有2个机房，分别是A与B，我就想查看A机房的，那么第一个参数就写'A'；
第二个参数是开始时间，如09；
第三个参数是结束时间，如13；
比如我想查看A机房早上9点到下午13点的超过10m流量，那么可以使用
python check_zabbix_network_traffic.py 'A' 09 13
默认输出目录是/tmp/，文件名格式是zabbix_network_traffic-当天-[开始时间-将结束时间],如zabbix_network_traffic-2015-08-19-[0-13].xls
效果为
报警组是zabbix里的报警组描述，主机就是ip信息，网卡名就是检测的网卡名，方向就是流入或者流出流量。
对于网卡名我在多描述一下，默认脚本里，网卡为em2、eth1、eth0
network_device=['em2','eth1','eth0']
为什么写这个呢，就是因为我这里主机网卡信息很混乱，比如openstack的云平台主机就一个网卡eth0，而centos5的网卡就是eth0于eth1，centos6 与centos7就都是em1与em2，就导致不通平台与不同系统，公网的设备名不一样，所以我这个脚本里做了判断，如果有em2网卡，就不检测eth1与eth0了，依次类推。
我写的目录就是为了快速查看依次流量，大家如果有其他的需求，可以自己根据需求改。