Add cpu_pinning and isolate_emulator_thread option

[brandboat]

related to harvester/harvester#6551, as title, add cpu_pinning and isolate_emulator_thread option.

Test Plan

Prerequisite

• rancher 2.9.1 (docker image)
• harvester single node (v1.4.0-rc, would be better to have 12CPUs as we will create VM with 8cpus in test plan)
  • > 8 CPUs and > 12GB mem
  • please enable cpu manager

Test steps

1. Build docker-machine-driver-harvester with this patch(i.e. PR)
2. Import harvester cluster to rancher (i.e. cluster-registration-url)
3. Go to rancher UI, deactivate harvester node driver
4. Copy the binary built in step 1. to path /opt/drivers/management-state/bin in rancher container

   docker cp docker-machine-driver-harvester <rancher_container_id>:/opt/drivers/management-state/bin/docker-machine-driver-harvester

5. In rancher container, delete harvester dynamic schema

   kubectl delete dynamicschemas.management.cattle.io/harvesterconfig

6. Go to rancher UI, activate harvester node driver
7. In rancher container, apply the following yaml files to deploy rke2 guest cluster to harvester. (make sure you fill in all fields start with <YOUR_... in the yaml examples.)
   • harvesterconfig.yaml

    apiVersion: rke-machine-config.cattle.io/v1
    cpuCount: "8"
    cpuPinning: true
    diskInfo: '{"disks":[{"imageName":"default/<YOUR_IMAGE_UNDER_DEFAULT_NS>","bootOrder":1,"size":40}]}'
    isolateEmulatorThread: true
    kind: HarvesterConfig
    memorySize: "12"
    metadata:
      name: test
      namespace: fleet-default
    networkInfo: '{"interfaces":[{"networkName":"default/<YOUR_VM_NERWORK_UNDER_DEFAULT_NS>","macAddress":""}]}'
    sshPort: "22"
    sshUser: ubuntu
    userData: I2Nsb3VkLWNvbmZpZwpwYWNrYWdlX3VwZGF0ZTogdHJ1ZQpwYWNrYWdlczoKICAtIHFlbXUtZ3Vlc3QtYWdlbnQKcnVuY21kOgogIC0gLSBzeXN0ZW1jdGwKICAgIC0gZW5hYmxlCiAgICAtICctLW5vdycKICAgIC0gcWVtdS1ndWVzdC1hZ2VudC5zZXJ2aWNlCg==
    vmNamespace: default

   • cluster.yaml

   apiVersion: provisioning.cattle.io/v1
    kind: Cluster
    metadata:
      name: foobar
      namespace: fleet-default
    spec:
      cloudCredentialSecretName: cattle-global-data:<YOUR_CLOUD_CREDENTIAL_SECRET_NAME>
      kubernetesVersion: v1.30.5+rke2r1
      rkeConfig:
        etcd:
          disableSnapshots: false
          snapshotRetention: 5
          snapshotScheduleCron: 0 */5 * * *
        machineGlobalConfig:
          cni: calico
          disable-kube-proxy: false
          etcd-expose-metrics: false
        machinePools:
          - name: pool1
            etcdRole: true
            controlPlaneRole: true
            workerRole: true
            quantity: 1
            unhealthyNodeTimeout: 0m
            machineConfigRef:
              kind: HarvesterConfig
              name: test
            drainBeforeDelete: true
            labels: {}
        machineSelectorConfig:
          - config:
              cloud-provider-name: harvester
              protect-kernel-defaults: false
        upgradeStrategy:
          controlPlaneConcurrency: '1'
          controlPlaneDrainOptions:
            deleteEmptyDirData: true
            disableEviction: false
            enabled: false
            force: false
            gracePeriod: -1
            ignoreDaemonSets: true
            skipWaitForDeleteTimeoutSeconds: 0
            timeout: 120
          workerConcurrency: '1'
          workerDrainOptions:
            deleteEmptyDirData: true
            disableEviction: false
            enabled: false
            force: false
            gracePeriod: -1
            ignoreDaemonSets: true
            skipWaitForDeleteTimeoutSeconds: 0
            timeout: 120

8. Go to harvester UI, navigate to Virtual Machines page, check the vm have the following attributes

    dedicatedCpuPlacement: true
    isolateEmulatorThread: true

   

[FrankYang0529]

LGTM. Thanks.

[bk201]

@FrankYang0529 do we need to ensure there is no such an upgrade path that upgrading from a Rancher with the fields to a Rancher without the new fields?

[FrankYang0529]

@FrankYang0529 do we need to ensure there is no such an upgrade path that upgrading from a Rancher with the fields to a Rancher without the new fields?

Yes, we need. If we only add it to latest rancher v2.9.x version, I think there is no way to downgrade to old rancher version.

[ibrokethecloud]

the changes look good, may be we just need to fine tune the usage message on the two flags.

[ibrokethecloud]

Suggested change

	Usage: "enable vm cpu pinning, make sure the the harvester cluster enable cpu manager in at least one node",
	Usage: "enable vm cpu pinning, please ensure the harvester cluster has cpu manager enabled in at least one node",

[ibrokethecloud]

Suggested change

	Usage: "enable vm isolatate emulator thread, note that enabling this feature will acquire one more cpu",
	Usage: "enable vm isolatated emulator thread",

[brandboat]

Thanks for the comments! I've addressed everything in the latest commit.

[ibrokethecloud]

lgtm. thanks.