Pretty much the same question so I’m exhuming this thread
As far as I was able to see on my RB1100AHx2, I can query CPU usage using:
.1.3.6.1.2.1.25.3.3.1.2.1 (cpu1)
.1.3.6.1.2.1.25.3.3.1.2.2 (cpu2)
.1.3.6.1.4.1.2021.11.10.0 (1min average?)
However none of this value reports what I can see in /system resources - Are cpu1 and 2 also 1min average?
And the last one is the average of both cpu on 1 min?
.1.3.6.1.2.1.25.3.3.1.2 is a standards based OID (coming from the host mgmt MIB).
The MIB specifies:
The average, over the last minute, of the percentage of time that this processor was not idle.
Implementations may approximate this one minute smoothing period if necessary.
Meaning the OID, according to the MIB, MUST return average CPU utilization for the last minute.
Previous to the change in v6.35, this OID returned IMMEDIATE CPU load at the time of the request.
This means MikroTik was incorrectly reporting CPU load as per the definition of the OID from the MIB.
So MikroTik changed .1.3.6.1.2.1.25.3.3.1.2 to properly report CPU load for the last minute.
A new OID was introduced, no not lose previous ability to poll current CPU load.
So .1.3.6.1.4.1.2021.11.10 will give you CPU load at the time you make the SNMP request.
On a RB1100AHx2 .1.3.6.1.2.1.25.3.3.1.2 exposes both CPUs (.1.3.6.1.2.1.25.3.3.1.2**.1** and .1.3.6.1.2.1.25.3.3.1.2**.2**)
However .1.3.6.1.4.1.2021.11.10 only exposes a single value. Would that be the average of both CPUs?
Also can someone please show where in the CLI we can access these OIDs (ie at what / run Print oid at)?
i never used to have an issue using the OID from /sys resource pr oid (or /sys resour cpu pr oid) but as of recent ROS versions / recent deployments the CLI is returning OID that give error.
bump - (mt support ?) We still dont have consistency pre-CPU OID change (ie around 6.35). How do we access (via snmp OID) the CPU load % shown in /system resource print (ie the avg of X cores, instant value). This is not a minor issue, its very important to graph / have access to CPU % data.
I appreciate the user reply below, but this is not working (tested on several MT devices, all return No such OID)
A new OID was introduced, no not lose previous ability to poll current CPU load.
So .1.3.6.1.4.1.2021.11.10 will give you CPU load at the time you make the SNMP request.
tks
EDIT: not even the 1m avg OID works, so mt pls, provide OID to get CPU % (and not on a per core per oid basis)
EDIT2: unless there was confuision in my email to MT support, it looks like they fully removed the ability to get any kind of CPU AVG (ie on multi-core RBs) around ros 6.35, and will consider adding it back in the future:
Thank you for contacting us.
We will consider adding OID for this value in future RouterOS releases.
Best regards,
Martins S.
There is no point in SNMP monitoring a single CPU core, of a multi-core RB ( ALL MT’s higher end RBs are Multi-core) - we need ability to get an average (back).
edit3; look like feature was removed inplace of per-core snmp monitoring only:
(below from MT support):
Currently in order to get CPU usage you have to get CPU core usage from System/Resource/CPU menu and calculate average value if you want to get total CPU usage.
Wanted to bump this back up. I’m unsure if the 1.3.6.1.4.1.2021.11.10 OID is working correctly. I see CPU loads of 5% under Resources on my CCR1036 but when I snmpwalk that OID I get an integer of zero. Now, weirdly, when I walk the OIDs for ALL the cores, they’re all around 1-4, some are zero. So I have no idea where the 5% number is coming from but the new OID shouldn’t be getting 0… Need to figure out a way to report a single data point in Zabbix for 36 cores…
