I have a appliance box that I need to monitor using the Spectrum net-snmp integration. I was searching in the doc's on how to set this up properly, but to my surprise I found nothing describing how to set this up. So the appliance box has an net-snmp which is preconfigured and I can see everything in Spectrum and the status. But as default, if status is fail, this does not create an alarm. So, I guess I have to create watches for this. This should be a pretty easy if I did know the attributes and how the MModel was designed in Spectrum. But as I can't find any doc's, its more complicated then it should be.
Example below, the appliance has a preconfigured processmonitor and I need to create an alarm if status is other than OK.
Any tip or guide to a doc I seem to have missed.
I suggest to poll the Mib for process table and get the OID from the result and then use it in watch. Altenate method is that if you have traps enabled on the end server we should be able to get them in Oneclick when the state changes.
Do you see the change in the Status value under monitored process when the process goes down ? Any unknown events on the model during this time?
Yes, the status values changes and of course I can create a watch that monitor this value. But interesting is that when creating a watch with name "prErrorFlag" which is the attribute that shows good/bad status for processes, then I get "A watch with this name already exists", which make me thing there are already some logic created for this...
Chapter 1: Introduction 9 in "Spectrum Host System Resource Management User Guide" PDF one can read a list of supported monitoring agents. Here Net-SNMP (UC Davis) is listed. But reading further in this doc, there is nothing more mentioned about Net-SNMP - which is very strange. There should have been mentioned how to proceed to activate montioring of Net-SNMP using Spectrum.
I was thinking that there maybe are watches created which can be activated using policies or alike... Anyone knows?
I believe it is talking about the duplicate watch name and not the attribute used in the expression. Can you give another watch name and use this attribute to see if it is alarmed?
We get an alarm when the number of process is not the desired one, via a Trap (see AlertMap file, where 0xffff2105 is a custum event of ours):
# UCD process thresh raised prIndex
18.104.22.168.4.1.2021.990.0.1 0xffff2105 22.214.171.124.4.1.2021.2.1.1(1,0)\
The Trap is sent by snmpd.
In snmpd conf we have:
# Event action and notification events
setEvent procFix prErrFix = 1
notificationEvent EventProcessKo .126.96.36.199.4.1.2021.990.0.1
notificationEvent EventProcessOk .188.8.131.52.4.1.2021.990.0.2
Hope this helps.