Monitoring SNMP service

[itzdan]

What is the best way to monitor the SNMP Service itself?

Is the best way to alarm to put a trigger on the SNMP service itself, alarming if it goes into an unkown state?

[kloczek]

SNMP service or more precisely SNMP agent is monitored over SNMPv2 MIB OIDs.
You can have look on my SNMPv2-MIB template https://github.com/kloczek/zabbix-te...MIB/SNMPv2-MIB

[itzdan]

If i'm monitoring from zabbix on the status of the agent. If the agent goes down, the response will be a no response/timeout value. Not a down value.

[kloczek]

If agent will go down you will be not able to collect even SNMPv2 MIB OIDs data
If you are planning/thinking about use this as indicator that exact device went down .. don't do that because it will give you in some scenarios false alarms.
For that should be used similar approach as I've implemented in my OS templates.
For example "host is down" trigger from my "OS Linux" template uses expression:

{OS Linux:icmpping.count(#10,0,"eq")}>9 and {OS Linux:agent.ping.nodata(15m)}=1 and {OS Linux:system.uptime.count(15m)}=0

Something like this should be implemented using SNMPv2-MIB template combined with ICMP (my "OS" templates are linked to ICMP ans "Service Zabbix Agent" templates providing enough data to use above logic).
Still I need to put somewhere some kind of documentation why this needs to be implemented exactly that way in case of active zabbix monitoring because it is not obvious .. (+/- use 15m as base time frame)

What is useful in SNMPv2 is related to SNMP timeouts issues. For example in exact metrics are counted those issues on agent side. With SNMPv2 metrics is possible to observe impact of using SNMP bulk queries (query more than one OID in single SNMP query).
SNMPv2-MIB::snmpPkts graph provided in my template provides kind of good observation deck of those effects.

Other useful metric provided in SNMPv2 is sysObjectID which contains list of MIBs over which SNMP agent exposes all other OIDs in whole SNMP tree.
On https://github.com/kloczek/zabbix-te...SNMP%20Devices you have short documentation about that part.
My long term idea about SNMP monitoring is do automatic detection using sysObjectID server content tan automatically add set of other MIB based templates basing on that output. I still have no idea how to implement and probably it will be not possible to do this without some changes in zabbix gut. This is why still if I have not complete idea how to implement that kind of self-assembly-template I'm still time to time I'm trying to spend some time on other per MIB based templates. When this senf-assembly process will be done it will be using already implemented per MIB templates. sysObjectID OID is already mapped to the SNMPvt-MIB template so base for that self-assembly already has been formed.

Nevertheless SNMPv2 MIB it is diagnostic MIB it does not describes any other useful things than exposing SNMP agent state metrics. On https://github.com/kloczek/zabbix-te...MIB/SNMPv2-MIB you can find link to http://www.oidview.com/mibs/0/SNMPv2-MIB.html which provides whole MIB description.

PS. Two weeks ago I've started my new contract under which I've been already asked to provide more sophisticated SNMP aware devices (mostly Cisco) so in next +2 weeks I'll be pushing probably more changes into https://github.com/kloczek/zabbix-te...tree/devel/MIB
As well withing next two weeks I'm going to push more "OS" and "Services" templates changes.
PS2. I'm not 100% sure is what you've been expecting as the answer on your question but what is above is literally what could be understood as "monitor the SNMP Service itself" topic.