[j-nsp] SNMP query delay

[Michael Skulsky]

Hi all,

 

I have a strange problem when retrieving interface statistics from IF-MIB
through SNMP. The router is an mx960. There are several interfaces, for
which the time needed to obtain the values is much higher than for others.
Compare:

 

xxx at test:~$ <mailto:os.arbor at test:~$>  time snmpget -v1 -c xxxxx -t 100
xx.xxx.xxx.xxx ifHCInOctets.108;

IF-MIB::ifHCInOctets.108 = Counter64: 0

real    0m0.070s

 

xxx at test:~$ <mailto:os.arbor at test:~$>  time snmpget -v1 -c xxxxx -t 100
xx.xxx.xxx.xxx ifHCInOctets.116;

IF-MIB::ifHCInOctets.116 = Counter64: 0

real    0m10.042s

 

All of the problematic interfaces are tunnel interfaces:

 

IF-MIB::ifType.116 = INTEGER: tunnel(131)

IF-MIB::ifType.117 = INTEGER: tunnel(131)

IF-MIB::ifType.118 = INTEGER: tunnel(131)

IF-MIB::ifType.119 = INTEGER: tunnel(131)

IF-MIB::ifType.120 = INTEGER: tunnel(131)

IF-MIB::ifType.121 = INTEGER: tunnel(131)

IF-MIB::ifType.122 = INTEGER: tunnel(131)

 

I don't know for sure which kind of tunnel this is (I'm troubleshooting
remotely a problem on customer side, and this is his router), and I'm not
even 100% sure that this is related to the root cause. Another common
property is that these interfaces report ifMtu = FFFFFFFF.

This delay is quite annoying, because it apparently slows down the SNMP
agent as a whole. All queries become sluggish, until the problematic one
completes, so my application cannot perform periodic statistic retrieval in
time. The problem is quite reproducible and occurs only on these interfaces
(so cannot be attributed to router load, etc).

 

Does anybody has any clues? Thanks in advance!

 

Michael