c8000v 17.15.3a in a Lab on KVM: vcode_pkt_PPE0 process CPU Hog

[dbillon]

Hi,

I setup an SDWAN Lab with several c8000v running IOS 17.15.3a running on QEMU 4.1.0 (EVE-NG Lab environment) with 4 vCPU and 8 GB of RAM for each c8000v.

CLI on the c8000v was very slow. I have spotted a process called vcode_pkt_PPE0 that consumes every cycle of every vCPU.

Is there a way to tune the setup of the c8000v to avoid this process to use all CPU cycles ?

Considering that it's just a Lab (no heavy traffic) I'm trying to switch to control-plane-heavy.

But are there other tweaks to optimize a c8000v in a Lab to avoid such a behavior (no need for high bandwidth / low usage of the data path) ? Specific parameters when launching the VM with QEMU ? Other configuration options ? Specific QEMU version to use ?

Thanks, Damien.

[Torbjørn]

I have c8000v nodes running fine on the same versions in my lab. I did however have some similar issues(can't remember if it was the exact same process) when I overprovisioned CPU on an older virtualisation host. Do you get the same issues when running fewer nodes? What CPU are you running?

[dbillon]

Actually the server is brand new Dell with 2 x Xeon Platinum 8548Y+ (48C/96T each)

[dbillon]

Thanks for the advice. Can you tell me more about the "system mode control-traffic" command.
I don't find this command on the c8000v