[cacti-user] Cacti + SNMP Time out

Hi All,

I am wondering what considerations need to be made when using Cacti to poll
Ubiquiti devices.

I have a number of devices filling my event logs with SNMP timeouts. After
troubleshooting, I found that it was due to the SNMP request buffer
rejecting any number of OIDs aggregated in a single request that are above
2. 

This device in question is a Bullet2, running firmware version 3.6. 

This was verified by lowering the "Maximum OID's Per Get Request" value to
2, and incrementing. With this value set to 3, 4, 5, 6, 7, 8, 9, 10, 11, (I
tried up to 11, but I logically assume from this it will also reject any
number greater than 11). 

See below logs.

5.15PM - Set to 11 max OIDs
5.20PM - Timeout.
5.20PM - Set to 10 max OIDs
5.26PM - Timeout.
5.26PM - Set to 5 max OIDs
07/08/2011 05:31:37 PM - SPINE: Poller[0] Host[3400] TH[1] DS[35344]
WARNING: SNMP timeout detected [500 ms], ignoring host '10.10.128.11' 
5.32PM - Max OIDs set to 4
07/08/2011 05:37:14 PM - SPINE: Poller[0] Host[3400] TH[1] DS[35341]
WARNING: SNMP timeout detected [500 ms], ignoring host '10.10.128.11' 
5.38PM - Max OID set to 3
07/08/2011 05:46:03 PM - SPINE: Poller[0] Host[3400] TH[1] DS[35342]
WARNING: SNMP timeout detected [500 ms], ignoring host '10.10.128.11' 
5.46PM - Set Max OIDs to 2. Monitored for 1 hour, no SNMP timeouts. 

Much to my horror, upon realizing that this morning, the SNMP logs are again
filled with SNMP timeouts. I have lowered the Maximum OIDs value from 2 to
1, and the SNMP timeouts have again stopped. Why did the SNMP timeouts stop
with 2 for so long, and then resume? 

07/12/2011 08:00:47 AM - SPINE: Poller[0] Host[3400] TH[1] DS[35350]
WARNING: SNMP timeout detected [500 ms], ignoring host '10.10.128.11'

Can anyone tell me how or what I should be doing to probe these devices to
see what the SNMP buffer is like? Is there a way to see a maximum number of
OIDs a particular device is capable of? Can this be changed? Polling 1 OID
for each of my hosts will effectively increase traffic, so is not an ideal
solution.

-- 

As an additional question, can somebody tell me how these OIDs are pulled? I
envisioned them to be aggregated into a single SNMPBulkGet request, however
as you can see below, this is not the case.

[templar@sandcastle ~]$ snmpbulkget -c ubnt -v 1 10.10.128.11
snmpbulkget: Cannot send V2 PDU on V1 session
[templar@sandcastle ~]$ snmpbulkget -c ubnt -v 2c 10.10.128.11
Timeout: No Response from 10.10.128.11
[templar@sandcastle ~]$ snmpwalk -c ubnt -v 2c 10.10.128.11
Timeout: No Response from 10.10.128.11
[templar@sandcastle ~]$ snmpwalk -c ubnt -v 1 10.10.128.11
SNMPv2-MIB::sysDescr.0 = STRING: Linux 2.4.27-ubnt0 #2 Mon Nov 29 11:08:58
EET 2010 mips
SNMPv2-MIB::sysObjectID.0 = OID: SNMPv2-SMI::enterprises.10002.1
DISMAN-EVENT-MIB::sysUpTimeInstance = Timeticks: (32185500) 3 days,
17:24:15.00

 

Regards,

 

--

 

Matthew T. Flannery
Network Operations Analyst