1. Documentation
×

Allocating specific CPUs for nodes in a cluster

When using the static CPU Manager policy, you can explicitly define a list of CPUs that are reserved for critical system processes on specific nodes. Reserving CPUs for critical system processes can help ensure cluster stability.

For example, on a system with 24 CPUs, you could reserve CPUs numbered 0 - 3 for the control plane allowing the compute nodes to use CPUs 4 - 23.

Reserving CPUs for nodes

You can explicitly define a list of CPUs that are reserved for critical system processes on specific nodes by creating a KubeletConfig custom resource (CR) to define the reservedSystemCPUs parameter. Reserving CPUs for critical system processes can help ensure cluster stability.

This list supersedes the CPUs that might be reserved by using the systemReserved parameter.

For more information on the systemReserved parameter, see "Allocating resources for nodes in an OKD cluster".

Prerequisites

  1. You have the label associated with the machine config pool (MCP) for the type of node you want to configure:

Procedure

  1. Create a YAML file for the KubeletConfig CR:

    apiVersion: machineconfiguration.openshift.io/v1
kind: KubeletConfig
metadata:
  name: set-reserved-cpus
spec:
  kubeletConfig:
    reservedSystemCPUs: "0,1,2,3"
  machineConfigPoolSelector:
    matchLabels:
      pools.operator.machineconfiguration.openshift.io/worker: ""
#...

    where:

    metadata.name

    Specifies a name for the CR.

    spec.kubeletConfig.reservedSystemCPUs

    Specifies the core IDs of the CPUs you want to reserve for the nodes associated with the MCP.

    spec.machineConfigPoolSelector.matchLabels

    Specifies the label from the MCP.

  2. Create the CR object:

    $ oc create -f <file_name>.yaml
Additional resources