Channels
extensions
rancher-extensions
supermicro-sixsq
ozt
os
one-point-x
rio
onlinemeetup
v16-v21-migration
events
onlinetraining
training-1220
training-0110
training-0124
k3s-contributor
submariner
storage
k3os
windows
ci-cd
theranchcast
rfed_ara
nederlands
ukranian
danish
training-0131
masterclass
training-0207
training-0214
terraform-controller
epinio
phillydotnet
swarm
rancher-wrangler
deutsch
russian
chinese
mesos
français
japanese
arm
longhorn-dev
terraform-provider-rke
service-mesh
azure
gcp
academy
random
elemental
espanol
lima
harvester-dev
portugues
kubernetes
kim
hypper
kubewarden
opni
logging
neuvector-security
rancher-setup
developer
office-hours
mexico
cabpr
rke
vsphere
longhorn-storage
k3s
k3d
terraform-provider-rancher2
fleet
amazon
harvester
rke2
s3gw
hobbyfarm
rancher-desktop
general
Title
w
witty-flower-16290
09/08/2022, 2:50 PMis there something obvious im missing but when i change the resource values in preferences for cpu usage etc my qemu process is ALWAYS locked to a single core...e
f
fast-garage-66093
09/08/2022, 3:47 PMHow do you determine the number of cores actually locked to qemu? And is this on Intel or Arm?
w
witty-flower-16290
09/08/2022, 3:59 PMim on arm i have a coworker o n intel, but their cpu usage will never exceed say 130% on mac, leading us to belive it's only using one core as if it was using all it woul be a far higher percentage with mac math
f
fast-garage-66093
09/08/2022, 5:27 PMIf I set the number of CPUs to 2, then I can see (In Activity Monitor) that the
qemu-system-x86_64Please let me know if you see something different.
w
witty-flower-16290
09/08/2022, 5:33 PMmyself and several other devs here are not seeing that
f
fast-garage-66093
09/08/2022, 5:34 PMYou have to provide more information. Did you reproduce it with the deployment I just sent you? How many cores does your machine have, are they hyperthreaded, how many did you allocate to RD, and did you use Activity Monitor to check the number.
I now did the same test on an M1 mini, and it has essentially the same results, except I get only 360% with 4 CPUs allocated, probably because of other stuff running on the machine.
Maybe you results are due to your workloads; maybe you are running just a single container, and the container is limited to one CPU or something. Then we would need to look at how you are running it, via kubernetes, or plain docker or nerdctl, and/or if the app isn't really multithreaded etc.
But my test shows that RD will use all allocated CPUs when you present it with multiple non-locking containers to make use of them.
I was now wondering if this might be related to the container engine, but I switched from containerd to moby and re-ran the test (on Intel), and still get the same results.