9800-40 IOSd high CPU utilization 1000 APs and AP SNMP requests

[jasonm002]

linux_iosd-image is getting its CPU resource allotment maxed out on a 9800-40 (e.g. from output of "show proc cpu history"  due to a large amount of SNMP. Concerned that this will cause problems with wireless RADIUS, is there any way to tune the CPU resources available to linux_iosd-image? The overall CPU utilization of the box is around 10% (show proc cpu platform sorted) but seems IOSd only gets a tiny slice of that on the 9800-40. Running 17.8.1 because of mDNS wired MAC filtering support. 

[balaji.bandi]

17.8.1 is latest code - so TAC may help you.

what happends when you disable SNMP ?

[jasonm002]

I can work around the issue by applying control-plane policer or filtering out SNMP from the host sending all the packets (it's an in-house monitoring system), just wondering if there was a way to tune resources available to IOSd because it doesn't seem to be granted much more than 10% of the total CPU resources. It's probably been like that in every version I would guess. People may not have noticed yet because I would guess most are not running over 1000 APs on a 9800-40.

[balaji.bandi]

Try your option see is that improves, not that i aware older code have that same issue.

but suggest to take TAC side view also good ( as mentioned this is new version of of code) - not many using, since you like to use because of the features you need it.

[Leo Laohoo]

When the CPU goes up again, post the complete output to the following commands: 

• sh platform resources
• sh platform software status con brief

[jasonm002 · Accepted solution]

I just worked around the issue for now by attaching an SNMP view that excludes ciscoMgmt.513 and enterprises.14179.2 which should be everything in the airespace and lwapp-ap MIBs. 

The issue should be reproducible on a controller with ~1000 APs if you have something requesting these SNMP OID names every minute or more frequently:

bsnAPIfOperStatus
bsnApIfNoOfUsers

cLLwappUpTime
cLApUpTime
cLApCpuCurrentUsage

these might also contribute but not sure

entSensorStatus
entPhySensorOperStatus

all of those had hundreds of request hits in the output of "show snmp stats oid" while the issue was occurring. 

i'm afraid I don't have the time for a TAC case at the moment but - issue has been worked around for now with the SNMP view. 

[Rich R]

Thanks - good to know.

1. That's very frequent to be polling those MIBs!
2. A lot of the Airespace MIBs have only been added in recent versions as they weren't supported at all in the earlier releases so a lot of this stuff is actually very new and not being used by many people.  Cisco would like everyone to use streaming telemetry instead on 9800.

Check out https://www.cisco.com/c/en/us/support/docs/wireless/catalyst-9800-series-wireless-controllers/217460-monitor-catalyst-9800-wlc-via-snmp-with.html
I've already reported a bug with the cpmCPUMemoryUsed & cpmCPUMemoryFree MIBs on HA-SSO pair - they stop working after switchover to chassis 2.