The point is that if I utilize the CPU on 1 guest, all the other guests see increase in CPU usage. I'm testing with stress-ng, but it's visible with normal workload as well. I really don't know if this is normal behaviour (but I don't think so, atleast i didn't notice stuff like this yet). I'm running the guests on deb9 with the distributed qemu and qemu-kvm packages installed. The host system has 24 cores and there are 32 vcores assigned. All vms have a 30% cfs (cgroup) quota on CPU usage, so the real amount of vcores assigned is 10 (9,6). The CPU usage on the host doesn't go above 15% if we don't do special stuff at that moment (idle state), the steal cpu usage is around 0,3-0,4 in the vms in the same idle state.
closed as too broad by Rui F Ribeiro, Jeff Schaller♦, Michael Homer, jimmij, Mr Shunz Mar 7 at 12:01
Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. Avoid asking multiple distinct questions at once. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.
Try pinning the CPUs on specific cores, so that they might not interfere: How to set KVM processors affinities with libvirt.
Other than that you may try to isolate some of the CPUs the guest is running on in order to make sure no user processes are executed on that core.
How to isolate CPUs in Ubuntu 18.04 with isolcpus | Linux kernel.
Also, it would be easier to troubleshoot if we had some more info, for example,
/proc/stat of the CPUs, or some