Zabbix 6.0.20 - SNMP V3 Low polling rate ( 1m reqst becomes 6 hours)

[Wolfsbane2k]

Hi.

I've got 3 APC UPS's that I'm monitoring using SNMP v3 that are generally being updated at a signficantly lower rate than desired: for example, Output Status (OID 1.3.6.1.4.1.318.1.1.1.4.1.1.0 upsBasicOutputStatus reference info (oidref.com)) , which is the "is it on Mains Power or not" is meant to be monitored every minute according to the update frequency, but Zabbix is monitoring it on a regular 6 hour interval - which means the triggers aren't being activated when they should be.

I'm using the inbuilt template "APC Smart-UPS SRT 5000 by SNMP" and have tried the generic one.

My system consists of less than 50 hosts, and my server config is set to 10 pollers, and Zabbix reports that the server is using less than 10% of internal procesess, 20% data collectors, 40% cache usage, and queue size of less than 200.

However, In an attempt to fix this, i've boosted numbers of pollers from 10 to 20, and seen no difference.

Other items from the same hosts are being monitored "correctly", such as "battery life remaining".

Any ideas appreciated!

[markosa]

Template seems to have :
So it does poll at configured interval but stores received values only when those are changed and every 6h

[Wolfsbane2k]

Thanks markosa. - that at least helps us identify why the graphs look weird!

I've gone into a wireshark trace and can see that Zabbix is actually polling that OID at 1m intervals.. which appears to aligns with your response and the manual around the use of that Discard preprocessor
This also explains why the Zabbix front end will continue to show "last check" as, currently, 5 + hours old, even though it's actually not - it's just not written to the DB.

Now I just need to understand why i didn't receive the alerts I was expecting when the power dropped....

[Wolfsbane2k]

Now I just need to understand why i didn't receive the alerts I was expecting when the power dropped....

Solved: numerous, repeated, short outages, none of which actually coincided with the 1m Zabbix polls, and there was something blocking the traps from being processed properly.