python snmp 自动化2-在python中使用snmp
python snmp 自动化2-在python中使用snmp#参考资料
为什么不使用pysnmp？
Pysnmp是一个纯python的snmp模块，不过我们的mib文件，大概有20%的，Pysnmp不能成功读取，报错如下：
# build-pysnmp-mib-o fsp150cm-sa.mib.txt fsp150cm-sa.mib
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
WARNING: empty MIBmodule name seen in smidump output at CM-SA-MIB
Traceback (mostrecent call last):
File "/usr/bin/libsmi2pysnmp", line5, in
pkg_resources.run_script('pysnmp==4.2.1','libsmi2pysnmp')
File "/usr/lib/python2.7/site-packages/setuptools-0.6c11-py2.7.egg/pkg_resources.py",line 489, in run_script
File"/usr/lib/python2.7/site-packages/setuptools-0.6c11-py2.7.egg/pkg_resources.py",line 1214, in run_script
File"/usr/lib/python2.7/site-packages/pysnmp-4.2.1-py2.7.egg/EGG-INFO/scripts/libsmi2pysnmp",line 435, in
File"/usr/lib/python2.7/site-packages/pysnmp-4.2.1-py2.7.egg/EGG-INFO/scripts/libsmi2pysnmp",line 147, in __genTypeDef
KeyError: 'syntax'
smidump -k -fpython fsp150cm-sa.mib | /usr/bin/libsmi2pysnmp fails
make sure you areusing libsmi version > 0.4.5 (or svn)
网上有篇netsnmp和pysnmp比较的评论：
This isn't so muchan answer as much as me sharing my experience testing these libraries.
I ran somesnmpwalk benchmarks on PySNMP 4.x and net-snmp using python'smultiprocessing.Pool. Benchmarks were pretty dirty (basically scaled up thethread pool until I didn't get any performance gains), so reader be-ware.Here's some observations:
netsnmp's pythonbindings:
* Could pull down 11958 oids/sec on thequad core test box.
* Didn't consume a lot of CPU. Seemed to bewaiting on network I/O most of the time (keep reading).
* Didn't support snmpbulkwalk,unfortunately. So this generates more network traffic and a single thread ofexecution tends to be slow.
* It has some annoying MIB look upbehavior. I had to specify 'ifName' instead of 'IF-MIB::ifName', which couldlead to some ambiguity. I also couldn't seem to find a way to control MIBlookups very well.
* Threading will not work well. Even if youwant one thread of execution, run it in a separate process so you don't starveother threads. This library is thread safe, but not thread friendly.
PySNMP4:
* Came in at 5560 oids/sec on the same box.
* Very CPU intensive. I attribute this tothe packet parsing being done in python.
* MIB lookups I thought were really nice.
* snmpwalks would leak some unrelated OIDs.For example, I'd walk IF-MIB::ifXTable and at the end I'd getIF-MIB::ifStackTable. IF-MIB dump.
* I'd almost certainly tailor a wrapper formy application instead of using this library directly. Specifically I'd wrapall the error handling to use Exceptions.
* I'm not a big fan of writing/readingasynchronous code, so I'd just ignore all the async bits and run big SNMPoperations in a separate process.
Overall, I'mreally kind of disappointed. There's really not a "best overall"library. Apart from the API and performance, PySNMP4 is great. Apart fromhaving some strange MIB/oid lookup handling behavior and not supporting manybulk operations, NetSNMP's python bindings were great.
可见pysnmp强在mib解析等方面，性能方面不能和netsnmp媲美。既然我们的mib pysnmp无法解析，只好放弃，期待pysnmp尽快开发出自己的好的mib解析器。只有选择netsnmp。
为什么不使用subprocess？
Subprocess会经常出现超时。
Python Get示例：
单个get：
print "Get ntpClientEnabled"
oid = netsnmp.Varbind('1.3.6.1.4.1','2544.1.12.2.1.10.1.0')
oidList = netsnmp.VarList(oid)
resultList = netsnmp.snmpget(oid,Version=2,DestHost='172.23.192.44',Community='private')
print resultList
以下是通过类的方法进行Get：
session =netsnmp.Session(Version=2,DestHost='172.23.192.44',Community='private')
oid =netsnmp.Varbind('1.3.6.1.4.1','2544.1.12.2.1.10.1.0',2,'INTEGER')
oidList = netsnmp.VarList(oid)
resultList = session.get(oidList)
print resultList
多个get：
print "Get ntpClientEnabled"
oid = netsnmp.Varbind('1.3.6.1.4.1','2544.1.12.2.1.10.1.0')
oid1 = netsnmp.Varbind('1.3.6.1.4.1.2544.1.12.8.1.1.1.5','1.1.1.1','15','INTEGER')
resultList = netsnmp.snmpget(oid,oid1,Version=2,DestHost='172.23.192.44',Community='private')
print resultList
以下是通过类的方法进行Get：
session =netsnmp.Session(Version=2,DestHost='172.23.192.44',Community='private')
oid2 =netsnmp.Varbind('1.3.6.1.4.1','2544.1.12.2.1.10.1.0',2,'INTEGER')
oid3 = netsnmp.Varbind('.1.3.6.1.4.1.2544.1.12.8.1.1.1.5','1.1.1.1',20,'INTEGER')
oidList = netsnmp.VarList(oid3,oid2)
resultList = session.get(oidList)
print resultList
注意类和函数的方式的区别。类方法始终要求传入的参数为oid列表，函数的方式则可以把单个或者多个oid作为参数传入，不需要oid列表。通过代码可以看到函数的方式其实调用类方法，为此，可能效率要比类方法低。
Python Set示例
调用netsnmp.snmpset一次set多个值：
print "SetntpClientEnabled"
oid =netsnmp.Varbind('1.3.6.1.4.1.2544.1.12.2.1.10.1','0','2','INTEGER')
oid1 =netsnmp.Varbind('1.3.6.1.4.1.2544.1.12.8.1.1.1.5','1.1.1.1','15','INTEGER')
resultList = netsnmp.snmpset(oid,oid1,Version=2,DestHost='172.23.192.44',Community='private')
print resultList
上面的1.3.6.1.4.1.前面是否加点号没有关系。Varbind的init方法如下，tag可以用tid，比如ntpClientEnabled，iid表示后面带的序号，比如0,1.1.1.1。val表示具体的值，type表示类型，更详细的参见python netsnmp库的README。Val值'2'和'15'可以不加引号，以为这里是整型，建议还是全部加上引号。
通过tcpdump抓包分析，snmpset多个值也发送了一次请求，因此多个set集合在一起发送，会明显提高效率。
class Varbind(object):
def __init__(self,tag=None, iid=None, val=None, type=None):
以下是通过类的方法进行set：
session =netsnmp.Session(Version=2,DestHost='172.23.192.44',Community='private')
oid2 =netsnmp.Varbind('1.3.6.1.4.1','2544.1.12.2.1.10.1.0',2,'INTEGER')
oid3 =netsnmp.Varbind('.1.3.6.1.4.1.2544.1.12.8.1.1.1.5','1.1.1.1',20,'INTEGER')
oidList =netsnmp.VarList(oid3,oid2)
resultList =session.set(oidList)
print resultList
：
Mib翻译
通过snmptranslate -Tso>toid.txt可以导出tid和oid的对应表。
通过snmptranslate –Td 可以获取oid的详细定义。
# snmptranslate-Td .1.3.6.1.4.1.2544.1.12.2.1.10.1
CM-SYSTEM-MIB::ntpClientEnabled
ntpClientEnabledOBJECT-TYPE
-- FROM CM-SYSTEM-MIB
-- TEXTUAL CONVENTION TruthValue
SYNTAX INTEGER {true(1), false(2)}
MAX-ACCESS read-write
STATUS current
DESCRIPTION"This allows to enable/disable the NTP client."
::= { iso(1)org(3) dod(6) internet(1) private(4) enterprises(1) advaMIB(2544) products(1)fsp150cm(12) cmSystemMIB(2) cmSystemObjects(1) cmTimeObjects(10) 1 }
为了提高可读性，我们需要在测试中使用true来表示1，false来表示2，并用ntpClientEnabled来表示.1.3.6.1.4.1.2544.1.12.2.1.10.1，遗憾的是，netsnmp的python库有如下限制“Access to theparsed MIB database is not yet implemented.”为此MIB的转换需要自行完成。为此拟定如下数据结构：
{ntpClientEnabled：.1.3.6.1.4.1.2544.1.12.2.1.10.1}
{ ntpClientEnabled:{ true:1, false:2}}
第2个字典只针对有别名的部分oid。
如果mib库不大，可以通过如下的脚本获取：
#!/usr/bin/python
# -*- coding: utf-8 -*-
# Function: Ssh to remote server
# Author: Andrew Xu
# CreateDate: 2012/02/28
import subprocess
import re
import multiprocessing
import pickle
import time
import random
def producer(sequence, input_p):
for itemin sequence:
# Putthe item on the queue
input_p.send(item)
# 消费者
def consumer(lock,pipe):
output_p,input_p = pipe
input_p.close() # 关闭管道输入口
whileTrue:
lock.acquire()
try:
time.sleep(1)
item= output_p.recv()
#管道读取失败的预防，重试一次
except Exception:
item= output_p.recv()
lock.release()
if item == None:
break
tid = item
oid = snmpDict
# 处理部分
try:
result= subprocess.check_output(['snmptranslate', '-Td',oid])
#subprocess失败的预防，重试一次
except Exception:
time.sleep(random.randint(1,8))
result= subprocess.check_output(['snmptranslate', '-Td',oid])
syntaxLine = syntax.search(result)
if syntaxLine:
syntaxLine= syntaxLine.group(1)
valueitems= items.findall(syntaxLine)
snmpValueDict= dict(valueitems)
else:
lock.acquire()
print(tid,"None")
lock.release()
continue
lock.acquire()
print(tid,dict(valueitems))
lock.release()
if __name__ == '__main__':
snmpDict =dict()
snmpValueDict= dict()
# 读取语法部分
syntax =re.compile(r'^ SYNTAX.*?\{(.*?)\}',re.MULTILINE)
# 读取字符串 和对应的值
items =re.compile('(\w*)\((\w*)\)')
f = open("snmpValueDict.txt",'w')
# 生成tid和oid对应的字典
for linein open("tid_oid.txt"):
tid, oid= line.strip().split(',')
snmpDict= oid
pickle.dump(snmpValueDict, open("snmpTidOid.txt","w"))
print"length of snmpDict:" + str(len(snmpDict))
# 进程数、创建管道，锁等
p_num = 10
process =[]
(output_p,input_p) = multiprocessing.Pipe()
lock =multiprocessing.Lock()
# 定义消费进程
for i inrange(p_num):
t=multiprocessing.Process(target=consumer,args=(lock,(output_p, input_p),))
t.daemon=True
process.append(t)
# 启动消费进程
for i in range(p_num):
process.start()
# 关闭输出管道，以往管道填充数据
output_p.close()
sequence =snmpDict.values() + *p_num
producer(snmpDict.keys(), input_p)
# 数据填充完毕，打开输入管道
input_p.close()
# 等待结束
for i inrange(p_num):
process.join()
pickle.dump(snmpValueDict, open("snmpValueDict.txt","w"))
print"length of snmpDict:" + str(len(snmpDict))
print"length of snmpValueDict:" + str(len(snmpValueDict))
以上也是一个多进程程序的演示，如果mib的oid超过1000条，建议采用如下脚本来提取：
#!/usr/bin/python
# -*- coding: utf-8 -*-
# Function: Ssh to remote server
# Author: Andrew Xu
# CreateDate: 2012/02/28
import re
import glob
import pickle
typeDict = dict()
snmpDict = dict()
snmpValueDict = dict()
# tid and oid dict
for line in open("tid_oid.txt"):
tid, oid =line.strip().split(',')
snmpDict = oid
pickle.dump(snmpDict,open("snmpTidOid.txt", "w"))
# type Dict
for mibFile inglob.glob(r"/home/share/andrew/soft/mib/110/MIBs/*.mib"):
files = open(mibFile).read()
text=re.sub("\s+"," ",files)
result =re.findall(r'(\w*)\s+OBJECT-TYPE SYNTAX\s+\w+\s\{(.*?\(\w\).*?)\}',text)
fordictName,values in result:
splitValues = re.findall('(\w+)\s*?\((\w+)\)',values)
typeDict = dict(splitValues)
pickle.dump(snmpDict, open("typeDict.txt","w"))
# snmp Value Dict
for mibFile inglob.glob(r"D:\soft\mib\110\MIBs\*.mib"):
printmibFile
files =open(mibFile).read()
text=re.sub("\s+"," ",files)
result =re.findall(r'\s(\w*?) ::= TEXTUAL-CONVENTION STATUS currentDESCRIPTION.*?SYNTAX\s+\w+\s\{(.*?\(\w\).*?)\}',text)
printresult
fordictName,values in result:
splitValues = re.findall('(\w+)\s*?\((\w+)\)',values)
snmpValueDict = dict(splitValues)
result =re.findall(r'\s(\w*?) ::= TEXTUAL-CONVENTION STATUS currentDESCRIPTION.*?SYNTAX\s+\w+\s\{(.*?\(\w\).*?)\}',text)
pickle.dump(snmpValueDict,open("snmpValueDict.txt", "w"))
上面的脚本需要根据mib文件格式的不同进行相应的调整。
现在就可以加载存储在文件中的字典，利用tid代替oid进行测试了。
#!/usr/bin/python
# -*- coding:utf-8 -*-
# Function: Ssh toremote server
# Author: Andrew Xu
# CreateDate:2012/02/28
import netsnmp
import re
import pickle
snmpTOidDict =pickle.load(open(r"snmpTidOid.txt"))
snmpValueDict =pickle.load(open(r"snmpValueDict.txt"))
print "SetntpClientEnabled"
printsnmpValueDict['ntpClientEnabled']
oid =netsnmp.Varbind(snmpTOidDict['ntpClientEnabled'], '0',snmpValueDict['ntpClientEnabled']['true'],'INTEGER')
oid1 = netsnmp.Varbind(snmpTOidDict['ecpaControlDuration'],'1.1.1.1','15','INTEGER')
resultList =netsnmp.snmpset(oid,oid1,Version=2,DestHost='172.23.192.44',Community='private')
print resultList
下面搞个复杂点的用例，在某产品中创建和删除保护组：
命令行执行如下：
# snmpset -v2c -cprivate 172.23.192.44 cmFacProtGroupSwitchMode.1.1.1.1i 1 cmFacProtGroupWorkPort.1.1.1.1 o 1.3.6.1.4.1.2544.1.12.4.1.7.1.1.1.1.1.1cmFacProtGroupProtPort.1.1.1.1 o 1.3.6.1.4.1.2544.1.12.4.1.7.1.1.1.1.1.2cmFacProtGroupRowStatus.1.1.1.1 i 4
CM-PROTECTION-MIB::cmFacProtGroupSwitchMode.1.1.1.1= INTEGER: oneplusone(1)
CM-PROTECTION-MIB::cmFacProtGroupWorkPort.1.1.1.1= OID: CM-FACILITY-MIB::cmEthernetNetPortIndex.1.1.1.1
CM-PROTECTION-MIB::cmFacProtGroupProtPort.1.1.1.1= OID: CM-FACILITY-MIB::cmEthernetNetPortIndex.1.1.1.2
CM-PROTECTION-MIB::cmFacProtGroupRowStatus.1.1.1.1= INTEGER: createAndGo(4)
# snmpset -v2c -cprivate172.23.192.44cmFacProtGroupRowStatus.1.1.1.1 i 6CM-PROTECTION-MIB::cmFacProtGroupRowStatus.1.1.1.1 = INTEGER: destroy(6)
通过python实现如下
#!/usr/bin/python
# -*- coding:utf-8 -*-
# Function: Ssh toremote server
# Author: Andrew Xu
# CreateDate:2012/02/28
import netsnmp
import re
import pickle
snmpTOidDict =pickle.load(open(r"snmpTidOid.txt"))
snmpValueDict =pickle.load(open(r"snmpValueDict.txt"))
printsnmpValueDict['cmFacProtGroupRowStatus']
oid =netsnmp.Varbind(snmpTOidDict['cmFacProtGroupSwitchMode'], '1.1.1.1',
snmpValueDict['cmFacProtGroupSwitchMode']['oneplusone'],'INTEGER')
oid1 = netsnmp.Varbind(snmpTOidDict['cmFacProtGroupWorkPort'],'1.1.1.1',
snmpTOidDict['cmEthernetNetPortIndex'] + '.1.1.1.1','OBJECTID')
oid2 =netsnmp.Varbind(snmpTOidDict['cmFacProtGroupProtPort'], '1.1.1.1',
snmpTOidDict['cmEthernetNetPortIndex']+ '.1.1.1.2','OBJECTID')
oid3 =netsnmp.Varbind(snmpTOidDict['cmFacProtGroupRowStatus'], '1.1.1.1',
snmpValueDict['cmFacProtGroupRowStatus']['createAndGo'],'INTEGER')
resultList =netsnmp.snmpset(oid, oid1,oid2,oid3,Version=2,DestHost='172.23.192.44',Community='private')
print resultList
oid3 =netsnmp.Varbind(snmpTOidDict['cmFacProtGroupRowStatus'], '1.1.1.1',
snmpValueDict['cmFacProtGroupRowStatus']['destroy'],'INTEGER')
resultList =netsnmp.snmpset( oid3,Version=2,DestHost='172.23.192.44',Community='private')
print resultList
Normal07.8 pt02falsefalsefalseEN-USZH-CNX-NONE/* Style Definitions */table.MsoNormalTable{mso-style-name:"Table Normal";mso-tstyle-rowband-size:0;mso-tstyle-colband-size:0;mso-style-noshow:yes;mso-style-priority:99;mso-style-parent:"";mso-padding-alt:0cm 5.4pt 0cm 5.4pt;mso-para-margin:0cm;mso-para-margin-bottom:.0001pt;mso-pagination:widow-orphan;font-size:10.5pt;mso-bidi-font-size:11.0pt;font-family:"Calibri","sans-serif";mso-ascii-font-family:Calibri;mso-ascii-theme-font:minor-latin;mso-hansi-font-family:Calibri;mso-hansi-theme-font:minor-latin;mso-bidi-font-family:"Times New Roman";mso-bidi-theme-font:minor-bidi;mso-font-kerning:1.0pt;}
当然上面这种调用方式，参数显得有点多，正式形成库时会再次封装。
性能
session =netsnmp.Session(Version=2,DestHost='172.23.192.44',Community='private')
oid =netsnmp.Varbind('1.3.6.1.4.1.2544.1.12','',2,'INTEGER')
oidList = netsnmp.VarList(oid)
resultList =session.walk(oidList)
print resultList
print len(resultList)
通过上述代码，执行time ./test.py 。walk的效率：耗时：5m38.919s oid个数：13385
平均秒只能获取约40个oid。和参考资料中测试得5560 oids/sec有百倍左右的差距。不过我们的设备已经吃到近90%的cpu，这部分有待以后继续研究优化。
页:
[1]