CPU Manager Example

This example flow uses a pre-built docker image based on cpuset-visualizer. Each container run here hosts an endpoint serving an up-to-the-second system representation generated by lstopo.

Setup

Start a local cluster with the static CPU manager policy enabled :

sudo PATH=$PATH \
  KUBELET_FLAGS="\
  --feature-gates=CPUManager=true \
  --cpu-manager-policy=static \
  --cpu-manager-reconcile-period=5s \
  --kube-reserved=cpu=500m" \
  ./hack/local-up-cluster.sh

Run pods

The example pods are in the /examples/cpu-manager directory of the Kubernetes source tree. Your ability to run all of the example pods simultaneously depends on how many CPUs are available on the test system.

The required CPUs for each example pod are listed below:

 POD              | CPUs
------------------|---------
 be.yaml          | >= 1 CPU
 shared.yaml      | >= 1 CPU
 exclusive-1.yaml | >= 2 CPU
 exclusive-2.yaml | >= 3 CPU
 exclusive-3.yaml | >= 4 CPU
 exclusive-4.yaml | >= 5 CPU

Example

Run a pod with a single container in the shared pool, and another pod with a single container in an exclusive cpuset with one CPU.

$ kubectl create -f examples/cpu-manager/shared.yaml
$ kubectl create -f examples/cpu-manager/exclusive-1.yaml

To list IP addresses of the pods running in the local cluster, do:

$ watch ./examples/cpu-manager/pod-ips
exclusive-1 http://172.17.0.4
shared http://172.17.0.3

1.7 KiB Raw Blame History

CPU Manager Example

Setup

Run pods

Example

Sample cpuset-visualizer output

1.7 KiB

Raw Blame History