解决CISCO SNMP CPU占用超80%

透过Cacti发现一台CISCO Route的CPU使用率都到90%以上了，赶快查看原因？

UM-WF#sh processes cpu sorted | include SNMP
CPU utilization for five seconds: 99%/8%; one minute: 99%; five minutes: 99%
 PID Runtime(uS)     Invoked      uSecs   5Sec   1Min   5Min TTY Process 
 233  4151587520      680981      37631 76.10% 77.15% 76.76%   0 SNMP ENGINE      
 231  3337772000     1302328       2562  8.79%  8.76%  8.74%   0 IP SNMP          
  93  1840900000     1222887       1505  3.63%  3.67%  3.72%   0 IP Input         

UM-WF#sh log
*Aug  9 10:08:34: %SNMP-3-INPUT_QFULL_ERR: Packet dropped due to input queue full
*Aug  9 10:08:34: %SNMP-3-INPUT_QFULL_ERR: Packet dropped due to input queue full
*Aug  9 10:09:00: %SNMP-3-INPUT_QFULL_ERR: Packet dropped due to input queue full

通过上述查证，证明了第一号战犯是SNMP ENGINE，奇怪的是过去运作几年了都没有出现问题过。于是，

UM-WF#sh run | begin snmp
snmp-server community public RO   //启用只读的snmp

只有这么一行，再加点限制

snmp-server host 172.x.x.x public //Specify hosts to receive SNMP notifications
snmp-server inform retries 0   //snmp v1的inform具有重传特写，默认是3

过一段时间观察还没有降下来，还是这么高。。。，再改，加个ACL上去，限定主机存取。

access-list 98 permit 172.x.x.x
access-list 98 deny   any
snmp-server community public RO 98

再来观察效果，终于降下来了。