×

You can create a different compute machine set to serve a specific purpose in your OKD cluster on Microsoft Azure. For example, you might create infrastructure machine sets and related machines so that you can move supporting workloads to the new machines.

You can use the advanced machine management and scaling capabilities only in clusters where the Machine API is operational. Clusters with user-provisioned infrastructure require additional validation and configuration to use the Machine API.

Clusters with the infrastructure platform type none cannot use the Machine API. This limitation applies even if the compute machines that are attached to the cluster are installed on a platform that supports the feature. This parameter cannot be changed after installation.

To view the platform type for your cluster, run the following command:

$ oc get infrastructure cluster -o jsonpath='{.status.platform}'

Sample YAML for a compute machine set custom resource on Azure

This sample YAML defines a compute machine set that runs in the 1 Microsoft Azure zone in a region and creates nodes that are labeled with node-role.kubernetes.io/<role>: "".

In this sample, <infrastructure_id> is the infrastructure ID label that is based on the cluster ID that you set when you provisioned the cluster, and <role> is the node label to add.

apiVersion: machine.openshift.io/v1beta1
kind: MachineSet
metadata:
  labels:
    machine.openshift.io/cluster-api-cluster: <infrastructure_id> (1)
    machine.openshift.io/cluster-api-machine-role: <role> (2)
    machine.openshift.io/cluster-api-machine-type: <role>
  name: <infrastructure_id>-<role>-<region> (3)
  namespace: openshift-machine-api
spec:
  replicas: 1
  selector:
    matchLabels:
      machine.openshift.io/cluster-api-cluster: <infrastructure_id>
      machine.openshift.io/cluster-api-machineset: <infrastructure_id>-<role>-<region>
  template:
    metadata:
      creationTimestamp: null
      labels:
        machine.openshift.io/cluster-api-cluster: <infrastructure_id>
        machine.openshift.io/cluster-api-machine-role: <role>
        machine.openshift.io/cluster-api-machine-type: <role>
        machine.openshift.io/cluster-api-machineset: <infrastructure_id>-<role>-<region>
    spec:
      metadata:
        creationTimestamp: null
        labels:
          machine.openshift.io/cluster-api-machineset: <machineset_name>
          node-role.kubernetes.io/<role>: ""
      providerSpec:
        value:
          apiVersion: azureproviderconfig.openshift.io/v1beta1
          credentialsSecret:
            name: azure-cloud-credentials
            namespace: openshift-machine-api
          image: (4)
            offer: ""
            publisher: ""
            resourceID: /resourceGroups/<infrastructure_id>-rg/providers/Microsoft.Compute/galleries/gallery_<infrastructure_id>/images/<infrastructure_id>-gen2/versions/latest (5)
            sku: ""
            version: ""
          internalLoadBalancer: ""
          kind: AzureMachineProviderSpec
          location: <region> (6)
          managedIdentity: <infrastructure_id>-identity
          metadata:
            creationTimestamp: null
          natRule: null
          networkResourceGroup: ""
          osDisk:
            diskSizeGB: 128
            managedDisk:
              storageAccountType: Premium_LRS
            osType: Linux
          publicIP: false
          publicLoadBalancer: ""
          resourceGroup: <infrastructure_id>-rg
          sshPrivateKey: ""
          sshPublicKey: ""
          tags:
            - name: <custom_tag_name> (7)
              value: <custom_tag_value>
          subnet: <infrastructure_id>-<role>-subnet
          userDataSecret:
            name: worker-user-data
          vmSize: Standard_D4s_v3
          vnet: <infrastructure_id>-vnet
          zone: "1" (8)
1 Specify the infrastructure ID that is based on the cluster ID that you set when you provisioned the cluster. If you have the OpenShift CLI installed, you can obtain the infrastructure ID by running the following command:
$ oc get -o jsonpath='{.status.infrastructureName}{"\n"}' infrastructure cluster

You can obtain the subnet by running the following command:

$  oc -n openshift-machine-api \
    -o jsonpath='{.spec.template.spec.providerSpec.value.subnet}{"\n"}' \
    get machineset/<infrastructure_id>-worker-centralus1

You can obtain the vnet by running the following command:

$  oc -n openshift-machine-api \
    -o jsonpath='{.spec.template.spec.providerSpec.value.vnet}{"\n"}' \
    get machineset/<infrastructure_id>-worker-centralus1
2 Specify the node label to add.
3 Specify the infrastructure ID, node label, and region.
4 Specify the image details for your compute machine set. If you want to use an Azure Marketplace image, see "Selecting an Azure Marketplace image".
5 Specify an image that is compatible with your instance type. The Hyper-V generation V2 images created by the installation program have a -gen2 suffix, while V1 images have the same name without the suffix.
6 Specify the region to place machines on.
7 Optional: Specify custom tags in your machine set. Provide the tag name in <custom_tag_name> field and the corresponding tag value in <custom_tag_value> field.
8 Specify the zone within your region to place machines on. Be sure that your region supports the zone that you specify.

Creating a compute machine set

In addition to the compute machine sets created by the installation program, you can create your own to dynamically manage the machine compute resources for specific workloads of your choice.

Prerequisites
  • Deploy an OKD cluster.

  • Install the OpenShift CLI (oc).

  • Log in to oc as a user with cluster-admin permission.

Procedure
  1. Create a new YAML file that contains the compute machine set custom resource (CR) sample and is named <file_name>.yaml.

    Ensure that you set the <clusterID> and <role> parameter values.

  2. Optional: If you are not sure which value to set for a specific field, you can check an existing compute machine set from your cluster.

    1. To list the compute machine sets in your cluster, run the following command:

      $ oc get machinesets -n openshift-machine-api
      Example output
      NAME                                DESIRED   CURRENT   READY   AVAILABLE   AGE
      agl030519-vplxk-worker-us-east-1a   1         1         1       1           55m
      agl030519-vplxk-worker-us-east-1b   1         1         1       1           55m
      agl030519-vplxk-worker-us-east-1c   1         1         1       1           55m
      agl030519-vplxk-worker-us-east-1d   0         0                             55m
      agl030519-vplxk-worker-us-east-1e   0         0                             55m
      agl030519-vplxk-worker-us-east-1f   0         0                             55m
    2. To view values of a specific compute machine set custom resource (CR), run the following command:

      $ oc get machineset <machineset_name> \
        -n openshift-machine-api -o yaml
      Example output
      apiVersion: machine.openshift.io/v1beta1
      kind: MachineSet
      metadata:
        labels:
          machine.openshift.io/cluster-api-cluster: <infrastructure_id> (1)
        name: <infrastructure_id>-<role> (2)
        namespace: openshift-machine-api
      spec:
        replicas: 1
        selector:
          matchLabels:
            machine.openshift.io/cluster-api-cluster: <infrastructure_id>
            machine.openshift.io/cluster-api-machineset: <infrastructure_id>-<role>
        template:
          metadata:
            labels:
              machine.openshift.io/cluster-api-cluster: <infrastructure_id>
              machine.openshift.io/cluster-api-machine-role: <role>
              machine.openshift.io/cluster-api-machine-type: <role>
              machine.openshift.io/cluster-api-machineset: <infrastructure_id>-<role>
          spec:
            providerSpec: (3)
              ...
      1 The cluster infrastructure ID.
      2 A default node label.

      For clusters that have user-provisioned infrastructure, a compute machine set can only create worker and infra type machines.

      3 The values in the <providerSpec> section of the compute machine set CR are platform-specific. For more information about <providerSpec> parameters in the CR, see the sample compute machine set CR configuration for your provider.
  3. Create a MachineSet CR by running the following command:

    $ oc create -f <file_name>.yaml
Verification
  • View the list of compute machine sets by running the following command:

    $ oc get machineset -n openshift-machine-api
    Example output
    NAME                                DESIRED   CURRENT   READY   AVAILABLE   AGE
    agl030519-vplxk-infra-us-east-1a    1         1         1       1           11m
    agl030519-vplxk-worker-us-east-1a   1         1         1       1           55m
    agl030519-vplxk-worker-us-east-1b   1         1         1       1           55m
    agl030519-vplxk-worker-us-east-1c   1         1         1       1           55m
    agl030519-vplxk-worker-us-east-1d   0         0                             55m
    agl030519-vplxk-worker-us-east-1e   0         0                             55m
    agl030519-vplxk-worker-us-east-1f   0         0                             55m

    When the new compute machine set is available, the DESIRED and CURRENT values match. If the compute machine set is not available, wait a few minutes and run the command again.

Selecting an Azure Marketplace image

You can create a machine set running on Azure that deploys machines that use the Azure Marketplace offering. To use this offering, you must first obtain the Azure Marketplace image. When obtaining your image, consider the following:

  • While the images are the same, the Azure Marketplace publisher is different depending on your region. If you are located in North America, specify redhat as the publisher. If you are located in EMEA, specify redhat-limited as the publisher.

  • The offer includes a rh-ocp-worker SKU and a rh-ocp-worker-gen1 SKU. The rh-ocp-worker SKU represents a Hyper-V generation version 2 VM image. The default instance types used in OKD are version 2 compatible. If you plan to use an instance type that is only version 1 compatible, use the image associated with the rh-ocp-worker-gen1 SKU. The rh-ocp-worker-gen1 SKU represents a Hyper-V version 1 VM image.

Installing images with the Azure marketplace is not supported on clusters with 64-bit ARM instances.

Prerequisites
  • You have installed the Azure CLI client (az).

  • Your Azure account is entitled for the offer and you have logged into this account with the Azure CLI client.

Procedure
  1. Display all of the available OKD images by running one of the following commands:

    • North America:

      $  az vm image list --all --offer rh-ocp-worker --publisher redhat -o table
      Example output
      Offer          Publisher       Sku                 Urn                                                             Version
      -------------  --------------  ------------------  --------------------------------------------------------------  --------------
      rh-ocp-worker  RedHat          rh-ocp-worker       RedHat:rh-ocp-worker:rh-ocpworker:4.8.2021122100               4.8.2021122100
      rh-ocp-worker  RedHat          rh-ocp-worker-gen1  RedHat:rh-ocp-worker:rh-ocp-worker-gen1:4.8.2021122100         4.8.2021122100
    • EMEA:

      $  az vm image list --all --offer rh-ocp-worker --publisher redhat-limited -o table
      Example output
      Offer          Publisher       Sku                 Urn                                                             Version
      -------------  --------------  ------------------  --------------------------------------------------------------  --------------
      rh-ocp-worker  redhat-limited  rh-ocp-worker       redhat-limited:rh-ocp-worker:rh-ocp-worker:4.8.2021122100       4.8.2021122100
      rh-ocp-worker  redhat-limited  rh-ocp-worker-gen1  redhat-limited:rh-ocp-worker:rh-ocp-worker-gen1:4.8.2021122100  4.8.2021122100

    Regardless of the version of OKD that you install, the correct version of the Azure Marketplace image to use is 4.8. If required, your VMs are automatically upgraded as part of the installation process.

  2. Inspect the image for your offer by running one of the following commands:

    • North America:

      $ az vm image show --urn redhat:rh-ocp-worker:rh-ocp-worker:<version>
    • EMEA:

      $ az vm image show --urn redhat-limited:rh-ocp-worker:rh-ocp-worker:<version>
  3. Review the terms of the offer by running one of the following commands:

    • North America:

      $ az vm image terms show --urn redhat:rh-ocp-worker:rh-ocp-worker:<version>
    • EMEA:

      $ az vm image terms show --urn redhat-limited:rh-ocp-worker:rh-ocp-worker:<version>
  4. Accept the terms of the offering by running one of the following commands:

    • North America:

      $ az vm image terms accept --urn redhat:rh-ocp-worker:rh-ocp-worker:<version>
    • EMEA:

      $ az vm image terms accept --urn redhat-limited:rh-ocp-worker:rh-ocp-worker:<version>
  5. Record the image details of your offer, specifically the values for publisher, offer, sku, and version.

  6. Add the following parameters to the providerSpec section of your machine set YAML file using the image details for your offer:

    Sample providerSpec image values for Azure Marketplace machines
    providerSpec:
      value:
        image:
          offer: rh-ocp-worker
          publisher: redhat
          resourceID: ""
          sku: rh-ocp-worker
          type: MarketplaceWithPlan
          version: 4.8.2021122100

Enabling Azure boot diagnostics

You can enable boot diagnostics on Azure machines that your machine set creates.

Prerequisites
  • Have an existing Microsoft Azure cluster.

Procedure
  • Add the diagnostics configuration that is applicable to your storage type to the providerSpec field in your machine set YAML file:

    • For an Azure Managed storage account:

      providerSpec:
        diagnostics:
          boot:
            storageAccountType: AzureManaged (1)
      1 Specifies an Azure Managed storage account.
    • For an Azure Unmanaged storage account:

      providerSpec:
        diagnostics:
          boot:
            storageAccountType: CustomerManaged (1)
            customerManaged:
              storageAccountURI: https://<storage-account>.blob.core.windows.net (2)
      1 Specifies an Azure Unmanaged storage account.
      2 Replace <storage-account> with the name of your storage account.

      Only the Azure Blob Storage data service is supported.

Verification
  • On the Microsoft Azure portal, review the Boot diagnostics page for a machine deployed by the machine set, and verify that you can see the serial logs for the machine.

Machine sets that deploy machines as Spot VMs

You can save on costs by creating a compute machine set running on Azure that deploys machines as non-guaranteed Spot VMs. Spot VMs utilize unused Azure capacity and are less expensive than standard VMs. You can use Spot VMs for workloads that can tolerate interruptions, such as batch or stateless, horizontally scalable workloads.

Azure can terminate a Spot VM at any time. Azure gives a 30-second warning to the user when an interruption occurs. OKD begins to remove the workloads from the affected instances when Azure issues the termination warning.

Interruptions can occur when using Spot VMs for the following reasons:

  • The instance price exceeds your maximum price

  • The supply of Spot VMs decreases

  • Azure needs capacity back

When Azure terminates an instance, a termination handler running on the Spot VM node deletes the machine resource. To satisfy the compute machine set replicas quantity, the compute machine set creates a machine that requests a Spot VM.

Creating Spot VMs by using compute machine sets

You can launch a Spot VM on Azure by adding spotVMOptions to your compute machine set YAML file.

Procedure
  • Add the following line under the providerSpec field:

    providerSpec:
      value:
        spotVMOptions: {}

    You can optionally set the spotVMOptions.maxPrice field to limit the cost of the Spot VM. For example you can set maxPrice: '0.98765'. If the maxPrice is set, this value is used as the hourly maximum spot price. If it is not set, the maximum price defaults to -1 and charges up to the standard VM price.

    Azure caps Spot VM prices at the standard price. Azure will not evict an instance due to pricing if the instance is set with the default maxPrice. However, an instance can still be evicted due to capacity restrictions.

It is strongly recommended to use the default standard VM price as the maxPrice value and to not set the maximum price for Spot VMs.

Machine sets that deploy machines on Ephemeral OS disks

You can create a compute machine set running on Azure that deploys machines on Ephemeral OS disks. Ephemeral OS disks use local VM capacity rather than remote Azure Storage. This configuration therefore incurs no additional cost and provides lower latency for reading, writing, and reimaging.

Additional resources

Creating machines on Ephemeral OS disks by using compute machine sets

You can launch machines on Ephemeral OS disks on Azure by editing your compute machine set YAML file.

Prerequisites
  • Have an existing Microsoft Azure cluster.

Procedure
  1. Edit the custom resource (CR) by running the following command:

    $ oc edit machineset <machine-set-name>

    where <machine-set-name> is the compute machine set that you want to provision machines on Ephemeral OS disks.

  2. Add the following to the providerSpec field:

    providerSpec:
      value:
        ...
        osDisk:
           ...
           diskSettings: (1)
             ephemeralStorageLocation: Local (1)
           cachingType: ReadOnly (1)
           managedDisk:
             storageAccountType: Standard_LRS (2)
           ...
    1 These lines enable the use of Ephemeral OS disks.
    2 Ephemeral OS disks are only supported for VMs or scale set instances that use the Standard LRS storage account type.

    The implementation of Ephemeral OS disk support in OKD only supports the CacheDisk placement type. Do not change the placement configuration setting.

  3. Create a compute machine set using the updated configuration:

    $ oc create -f <machine-set-config>.yaml
Verification
  • On the Microsoft Azure portal, review the Overview page for a machine deployed by the compute machine set, and verify that the Ephemeral OS disk field is set to OS cache placement.

Machine sets that deploy machines with ultra disks as data disks

You can create a machine set running on Azure that deploys machines with ultra disks. Ultra disks are high-performance storage that are intended for use with the most demanding data workloads.

You can also create a persistent volume claim (PVC) that dynamically binds to a storage class backed by Azure ultra disks and mounts them to pods.

Data disks do not support the ability to specify disk throughput or disk IOPS. You can configure these properties by using PVCs.

Creating machines with ultra disks by using machine sets

You can deploy machines with ultra disks on Azure by editing your machine set YAML file.

Prerequisites
  • Have an existing Microsoft Azure cluster.

Procedure
  1. Create a custom secret in the openshift-machine-api namespace using the worker data secret by running the following command:

    $ oc -n openshift-machine-api \
    get secret <role>-user-data \ (1)
    --template='{{index .data.userData | base64decode}}' | jq > userData.txt (2)
    1 Replace <role> with worker.
    2 Specify userData.txt as the name of the new custom secret.
  2. In a text editor, open the userData.txt file and locate the final } character in the file.

    1. On the immediately preceding line, add a ,.

    2. Create a new line after the , and add the following configuration details:

      "storage": {
        "disks": [ (1)
          {
            "device": "/dev/disk/azure/scsi1/lun0", (2)
            "partitions": [ (3)
              {
                "label": "lun0p1", (4)
                "sizeMiB": 1024, (5)
                "startMiB": 0
              }
            ]
          }
        ],
        "filesystems": [ (6)
          {
            "device": "/dev/disk/by-partlabel/lun0p1",
            "format": "xfs",
            "path": "/var/lib/lun0p1"
          }
        ]
      },
      "systemd": {
        "units": [ (7)
          {
            "contents": "[Unit]\nBefore=local-fs.target\n[Mount]\nWhere=/var/lib/lun0p1\nWhat=/dev/disk/by-partlabel/lun0p1\nOptions=defaults,pquota\n[Install]\nWantedBy=local-fs.target\n", (8)
            "enabled": true,
            "name": "var-lib-lun0p1.mount"
          }
        ]
      }
      1 The configuration details for the disk that you want to attach to a node as an ultra disk.
      2 Specify the lun value that is defined in the dataDisks stanza of the machine set you are using. For example, if the machine set contains lun: 0, specify lun0. You can initialize multiple data disks by specifying multiple "disks" entries in this configuration file. If you specify multiple "disks" entries, ensure that the lun value for each matches the value in the machine set.
      3 The configuration details for a new partition on the disk.
      4 Specify a label for the partition. You might find it helpful to use hierarchical names, such as lun0p1 for the first partition of lun0.
      5 Specify the total size in MiB of the partition.
      6 Specify the filesystem to use when formatting a partition. Use the partition label to specify the partition.
      7 Specify a systemd unit to mount the partition at boot. Use the partition label to specify the partition. You can create multiple partitions by specifying multiple "partitions" entries in this configuration file. If you specify multiple "partitions" entries, you must specify a systemd unit for each.
      8 For Where, specify the value of storage.filesystems.path. For What, specify the value of storage.filesystems.device.
  3. Extract the disabling template value to a file called disableTemplating.txt by running the following command:

    $ oc -n openshift-machine-api get secret <role>-user-data \ (1)
    --template='{{index .data.disableTemplating | base64decode}}' | jq > disableTemplating.txt
    1 Replace <role> with worker.
  4. Combine the userData.txt file and disableTemplating.txt file to create a data secret file by running the following command:

    $ oc -n openshift-machine-api create secret generic <role>-user-data-x5 \ (1)
    --from-file=userData=userData.txt \
    --from-file=disableTemplating=disableTemplating.txt
    1 For <role>-user-data-x5, specify the name of the secret. Replace <role> with worker.
  5. Copy an existing Azure MachineSet custom resource (CR) and edit it by running the following command:

    $ oc edit machineset <machine-set-name>

    where <machine-set-name> is the machine set that you want to provision machines with ultra disks.

  6. Add the following lines in the positions indicated:

    apiVersion: machine.openshift.io/v1beta1
    kind: MachineSet
    spec:
      template:
        spec:
          metadata:
            labels:
              disk: ultrassd (1)
          providerSpec:
            value:
              ultraSSDCapability: Enabled (2)
              dataDisks: (2)
              - nameSuffix: ultrassd
                lun: 0
                diskSizeGB: 4
                deletionPolicy: Delete
                cachingType: None
                managedDisk:
                  storageAccountType: UltraSSD_LRS
              userDataSecret:
                name: <role>-user-data-x5 (3)
    1 Specify a label to use to select a node that is created by this machine set. This procedure uses disk.ultrassd for this value.
    2 These lines enable the use of ultra disks. For dataDisks, include the entire stanza.
    3 Specify the user data secret created earlier. Replace <role> with worker.
  7. Create a machine set using the updated configuration by running the following command:

    $ oc create -f <machine-set-name>.yaml
Verification
  1. Validate that the machines are created by running the following command:

    $ oc get machines

    The machines should be in the Running state.

  2. For a machine that is running and has a node attached, validate the partition by running the following command:

    $ oc debug node/<node-name> -- chroot /host lsblk

    In this command, oc debug node/<node-name> starts a debugging shell on the node <node-name> and passes a command with --. The passed command chroot /host provides access to the underlying host OS binaries, and lsblk shows the block devices that are attached to the host OS machine.

Next steps
  • To use an ultra disk from within a pod, create a workload that uses the mount point. Create a YAML file similar to the following example:

    apiVersion: v1
    kind: Pod
    metadata:
      name: ssd-benchmark1
    spec:
      containers:
      - name: ssd-benchmark1
        image: nginx
        ports:
          - containerPort: 80
            name: "http-server"
        volumeMounts:
        - name: lun0p1
          mountPath: "/tmp"
      volumes:
        - name: lun0p1
          hostPath:
            path: /var/lib/lun0p1
            type: DirectoryOrCreate
      nodeSelector:
        disktype: ultrassd

Troubleshooting resources for machine sets that enable ultra disks

Use the information in this section to understand and recover from issues you might encounter.

Incorrect ultra disk configuration

If an incorrect configuration of the ultraSSDCapability parameter is specified in the machine set, the machine provisioning fails.

For example, if the ultraSSDCapability parameter is set to Disabled, but an ultra disk is specified in the dataDisks parameter, the following error message appears:

StorageAccountType UltraSSD_LRS can be used only when additionalCapabilities.ultraSSDEnabled is set.
  • To resolve this issue, verify that your machine set configuration is correct.

Unsupported disk parameters

If a region, availability zone, or instance size that is not compatible with ultra disks is specified in the machine set, the machine provisioning fails. Check the logs for the following error message:

failed to create vm <machine_name>: failure sending request for machine <machine_name>: cannot create vm: compute.VirtualMachinesClient#CreateOrUpdate: Failure sending request: StatusCode=400 -- Original Error: Code="BadRequest" Message="Storage Account type 'UltraSSD_LRS' is not supported <more_information_about_why>."
  • To resolve this issue, verify that you are using this feature in a supported environment and that your machine set configuration is correct.

Unable to delete disks

If the deletion of ultra disks as data disks is not working as expected, the machines are deleted and the data disks are orphaned. You must delete the orphaned disks manually if desired.

Enabling customer-managed encryption keys for a machine set

You can supply an encryption key to Azure to encrypt data on managed disks at rest. You can enable server-side encryption with customer-managed keys by using the Machine API.

An Azure Key Vault, a disk encryption set, and an encryption key are required to use a customer-managed key. The disk encryption set must be in a resource group where the Cloud Credential Operator (CCO) has granted permissions. If not, an additional reader role is required to be granted on the disk encryption set.

Procedure
  • Configure the disk encryption set under the providerSpec field in your machine set YAML file. For example:

    providerSpec:
      value:
        osDisk:
          diskSizeGB: 128
          managedDisk:
            diskEncryptionSet:
              id: /subscriptions/<subscription_id>/resourceGroups/<resource_group_name>/providers/Microsoft.Compute/diskEncryptionSets/<disk_encryption_set_name>
            storageAccountType: Premium_LRS

Accelerated Networking for Microsoft Azure VMs

Accelerated Networking uses single root I/O virtualization (SR-IOV) to provide Microsoft Azure VMs with a more direct path to the switch. This enhances network performance. This feature can be enabled during or after installation.

Limitations

Consider the following limitations when deciding whether to use Accelerated Networking:

  • Accelerated Networking is only supported on clusters where the Machine API is operational.

  • Although the minimum requirement for an Azure worker node is two vCPUs, Accelerated Networking requires an Azure VM size that includes at least four vCPUs. To satisfy this requirement, you can change the value of vmSize in your machine set. For information about Azure VM sizes, see Microsoft Azure documentation.

  • When this feature is enabled on an existing Azure cluster, only newly provisioned nodes are affected. Currently running nodes are not reconciled. To enable the feature on all nodes, you must replace each existing machine. This can be done for each machine individually, or by scaling the replicas down to zero, and then scaling back up to your desired number of replicas.

Adding a GPU node to an existing OKD cluster

You can copy and modify a default compute machine set configuration to create a GPU-enabled machine set and machines for the Azure cloud provider.

The following table lists the validated instance types:

vmSize NVIDIA GPU accelerator Maximum number of GPUs Architecture

Standard_NC24s_v3

V100

4

x86

Standard_NC4as_T4_v3

T4

1

x86

ND A100 v4

A100

8

x86

By default, Azure subscriptions do not have a quota for the Azure instance types with GPU. Customers have to request a quota increase for the Azure instance families listed above.

Procedure
  1. View the machines and machine sets that exist in the openshift-machine-api namespace by running the following command. Each compute machine set is associated with a different availability zone within the Azure region. The installer automatically load balances compute machines across availability zones.

    $ oc get machineset -n openshift-machine-api
    Example output
    NAME                              DESIRED   CURRENT   READY   AVAILABLE   AGE
    myclustername-worker-centralus1   1         1         1       1           6h9m
    myclustername-worker-centralus2   1         1         1       1           6h9m
    myclustername-worker-centralus3   1         1         1       1           6h9m
  2. Make a copy of one of the existing compute MachineSet definitions and output the result to a YAML file by running the following command. This will be the basis for the GPU-enabled compute machine set definition.

    $ oc get machineset -n openshift-machine-api myclustername-worker-centralus1 -o yaml > machineset-azure.yaml
  3. View the content of the machineset:

    $ cat machineset-azure.yaml
    Example machineset-azure.yaml file
    apiVersion: machine.openshift.io/v1beta1
    kind: MachineSet
    metadata:
      annotations:
        machine.openshift.io/GPU: "0"
        machine.openshift.io/memoryMb: "16384"
        machine.openshift.io/vCPU: "4"
      creationTimestamp: "2023-02-06T14:08:19Z"
      generation: 1
      labels:
        machine.openshift.io/cluster-api-cluster: myclustername
        machine.openshift.io/cluster-api-machine-role: worker
        machine.openshift.io/cluster-api-machine-type: worker
      name: myclustername-worker-centralus1
      namespace: openshift-machine-api
      resourceVersion: "23601"
      uid: acd56e0c-7612-473a-ae37-8704f34b80de
    spec:
      replicas: 1
      selector:
        matchLabels:
          machine.openshift.io/cluster-api-cluster: myclustername
          machine.openshift.io/cluster-api-machineset: myclustername-worker-centralus1
      template:
        metadata:
          labels:
            machine.openshift.io/cluster-api-cluster: myclustername
            machine.openshift.io/cluster-api-machine-role: worker
            machine.openshift.io/cluster-api-machine-type: worker
            machine.openshift.io/cluster-api-machineset: myclustername-worker-centralus1
        spec:
          lifecycleHooks: {}
          metadata: {}
          providerSpec:
            value:
              acceleratedNetworking: true
              apiVersion: machine.openshift.io/v1beta1
              credentialsSecret:
                name: azure-cloud-credentials
                namespace: openshift-machine-api
              diagnostics: {}
              image:
                offer: ""
                publisher: ""
                resourceID: /resourceGroups/myclustername-rg/providers/Microsoft.Compute/galleries/gallery_myclustername_n6n4r/images/myclustername-gen2/versions/latest
                sku: ""
                version: ""
              kind: AzureMachineProviderSpec
              location: centralus
              managedIdentity: myclustername-identity
              metadata:
                creationTimestamp: null
              networkResourceGroup: myclustername-rg
              osDisk:
                diskSettings: {}
                diskSizeGB: 128
                managedDisk:
                  storageAccountType: Premium_LRS
                osType: Linux
              publicIP: false
              publicLoadBalancer: myclustername
              resourceGroup: myclustername-rg
              spotVMOptions: {}
              subnet: myclustername-worker-subnet
              userDataSecret:
                name: worker-user-data
              vmSize: Standard_D4s_v3
              vnet: myclustername-vnet
              zone: "1"
    status:
      availableReplicas: 1
      fullyLabeledReplicas: 1
      observedGeneration: 1
      readyReplicas: 1
      replicas: 1
  4. Make a copy of the machineset-azure.yaml file by running the following command:

    $ cp machineset-azure.yaml machineset-azure-gpu.yaml
  5. Update the following fields in machineset-azure-gpu.yaml:

    • Change .metadata.name to a name containing gpu.

    • Change .spec.selector.matchLabels["machine.openshift.io/cluster-api-machineset"] to match the new .metadata.name.

    • Change .spec.template.metadata.labels["machine.openshift.io/cluster-api-machineset"] to match the new .metadata.name.

    • Change .spec.template.spec.providerSpec.value.vmSize to Standard_NC4as_T4_v3.

      Example machineset-azure-gpu.yaml file
      apiVersion: machine.openshift.io/v1beta1
      kind: MachineSet
      metadata:
        annotations:
          machine.openshift.io/GPU: "1"
          machine.openshift.io/memoryMb: "28672"
          machine.openshift.io/vCPU: "4"
        creationTimestamp: "2023-02-06T20:27:12Z"
        generation: 1
        labels:
          machine.openshift.io/cluster-api-cluster: myclustername
          machine.openshift.io/cluster-api-machine-role: worker
          machine.openshift.io/cluster-api-machine-type: worker
        name: myclustername-nc4ast4-gpu-worker-centralus1
        namespace: openshift-machine-api
        resourceVersion: "166285"
        uid: 4eedce7f-6a57-4abe-b529-031140f02ffa
      spec:
        replicas: 1
        selector:
          matchLabels:
            machine.openshift.io/cluster-api-cluster: myclustername
            machine.openshift.io/cluster-api-machineset: myclustername-nc4ast4-gpu-worker-centralus1
        template:
          metadata:
            labels:
              machine.openshift.io/cluster-api-cluster: myclustername
              machine.openshift.io/cluster-api-machine-role: worker
              machine.openshift.io/cluster-api-machine-type: worker
              machine.openshift.io/cluster-api-machineset: myclustername-nc4ast4-gpu-worker-centralus1
          spec:
            lifecycleHooks: {}
            metadata: {}
            providerSpec:
              value:
                acceleratedNetworking: true
                apiVersion: machine.openshift.io/v1beta1
                credentialsSecret:
                  name: azure-cloud-credentials
                  namespace: openshift-machine-api
                diagnostics: {}
                image:
                  offer: ""
                  publisher: ""
                  resourceID: /resourceGroups/myclustername-rg/providers/Microsoft.Compute/galleries/gallery_myclustername_n6n4r/images/myclustername-gen2/versions/latest
                  sku: ""
                  version: ""
                kind: AzureMachineProviderSpec
                location: centralus
                managedIdentity: myclustername-identity
                metadata:
                  creationTimestamp: null
                networkResourceGroup: myclustername-rg
                osDisk:
                  diskSettings: {}
                  diskSizeGB: 128
                  managedDisk:
                    storageAccountType: Premium_LRS
                  osType: Linux
                publicIP: false
                publicLoadBalancer: myclustername
                resourceGroup: myclustername-rg
                spotVMOptions: {}
                subnet: myclustername-worker-subnet
                userDataSecret:
                  name: worker-user-data
                vmSize: Standard_NC4as_T4_v3
                vnet: myclustername-vnet
                zone: "1"
      status:
        availableReplicas: 1
        fullyLabeledReplicas: 1
        observedGeneration: 1
        readyReplicas: 1
        replicas: 1
  6. To verify your changes, perform a diff of the original compute definition and the new GPU-enabled node definition by running the following command:

    $ diff machineset-azure.yaml machineset-azure-gpu.yaml
    Example output
    14c14
    <   name: myclustername-worker-centralus1
    ---
    >   name: myclustername-nc4ast4-gpu-worker-centralus1
    23c23
    <       machine.openshift.io/cluster-api-machineset: myclustername-worker-centralus1
    ---
    >       machine.openshift.io/cluster-api-machineset: myclustername-nc4ast4-gpu-worker-centralus1
    30c30
    <         machine.openshift.io/cluster-api-machineset: myclustername-worker-centralus1
    ---
    >         machine.openshift.io/cluster-api-machineset: myclustername-nc4ast4-gpu-worker-centralus1
    67c67
    <           vmSize: Standard_D4s_v3
    ---
    >           vmSize: Standard_NC4as_T4_v3
  7. Create the GPU-enabled compute machine set from the definition file by running the following command:

    $ oc create -f machineset-azure-gpu.yaml
    Example output
    machineset.machine.openshift.io/myclustername-nc4ast4-gpu-worker-centralus1 created
  8. View the machines and machine sets that exist in the openshift-machine-api namespace by running the following command. Each compute machine set is associated with a different availability zone within the Azure region. The installer automatically load balances compute machines across availability zones.

    $ oc get machineset -n openshift-machine-api
    Example output
    NAME                                               DESIRED   CURRENT   READY   AVAILABLE   AGE
    clustername-n6n4r-nc4ast4-gpu-worker-centralus1    1         1         1       1           122m
    clustername-n6n4r-worker-centralus1                1         1         1       1           8h
    clustername-n6n4r-worker-centralus2                1         1         1       1           8h
    clustername-n6n4r-worker-centralus3                1         1         1       1           8h
  9. View the machines that exist in the openshift-machine-api namespace by running the following command. You can only configure one compute machine per set, although you can scale a compute machine set to add a node in a particular region and zone.

    $ oc get machines -n openshift-machine-api
    Example output
    NAME                                                PHASE     TYPE                   REGION      ZONE   AGE
    myclustername-master-0                              Running   Standard_D8s_v3        centralus   2      6h40m
    myclustername-master-1                              Running   Standard_D8s_v3        centralus   1      6h40m
    myclustername-master-2                              Running   Standard_D8s_v3        centralus   3      6h40m
    myclustername-nc4ast4-gpu-worker-centralus1-w9bqn   Running      centralus   1      21m
    myclustername-worker-centralus1-rbh6b               Running   Standard_D4s_v3        centralus   1      6h38m
    myclustername-worker-centralus2-dbz7w               Running   Standard_D4s_v3        centralus   2      6h38m
    myclustername-worker-centralus3-p9b8c               Running   Standard_D4s_v3        centralus   3      6h38m
  10. View the existing nodes, machines, and machine sets by running the following command. Note that each node is an instance of a machine definition with a specific Azure region and OKD role.

    $ oc get nodes
    Example output
    NAME                                                STATUS   ROLES                  AGE     VERSION
    myclustername-master-0                              Ready    control-plane,master   6h39m   v1.25.4+a34b9e9
    myclustername-master-1                              Ready    control-plane,master   6h41m   v1.25.4+a34b9e9
    myclustername-master-2                              Ready    control-plane,master   6h39m   v1.25.4+a34b9e9
    myclustername-nc4ast4-gpu-worker-centralus1-w9bqn   Ready    worker                 14m     v1.25.4+a34b9e9
    myclustername-worker-centralus1-rbh6b               Ready    worker                 6h29m   v1.25.4+a34b9e9
    myclustername-worker-centralus2-dbz7w               Ready    worker                 6h29m   v1.25.4+a34b9e9
    myclustername-worker-centralus3-p9b8c               Ready    worker                 6h31m   v1.25.4+a34b9e9
  11. View the list of compute machine sets:

    $ oc get machineset -n openshift-machine-api
    Example output
    NAME                                   DESIRED   CURRENT   READY   AVAILABLE   AGE
    myclustername-worker-centralus1        1         1         1       1           8h
    myclustername-worker-centralus2        1         1         1       1           8h
    myclustername-worker-centralus3        1         1         1       1           8h
  12. Create the GPU-enabled compute machine set from the definition file by running the following command:

    $ oc create -f machineset-azure-gpu.yaml
  13. View the list of compute machine sets:

    oc get machineset -n openshift-machine-api
    Example output
    NAME                                          DESIRED   CURRENT   READY   AVAILABLE   AGE
    myclustername-nc4ast4-gpu-worker-centralus1   1         1         1       1           121m
    myclustername-worker-centralus1               1         1         1       1           8h
    myclustername-worker-centralus2               1         1         1       1           8h
    myclustername-worker-centralus3               1         1         1       1           8h
Verification
  1. View the machine set you created by running the following command:

    $ oc get machineset -n openshift-machine-api | grep gpu

    The MachineSet replica count is set to 1 so a new Machine object is created automatically.

    Example output
    myclustername-nc4ast4-gpu-worker-centralus1   1         1         1       1           121m
  2. View the Machine object that the machine set created by running the following command:

    $ oc -n openshift-machine-api get machines | grep gpu
    Example output
    myclustername-nc4ast4-gpu-worker-centralus1-w9bqn   Running   Standard_NC4as_T4_v3   centralus   1      21m

There is no need to specify a namespace for the node. The node definition is cluster scoped.

Deploying the Node Feature Discovery Operator

After the GPU-enabled node is created, you need to discover the GPU-enabled node so it can be scheduled. To do this, install the Node Feature Discovery (NFD) Operator. The NFD Operator identifies hardware device features in nodes. It solves the general problem of identifying and cataloging hardware resources in the infrastructure nodes so they can be made available to OKD.

Procedure
  1. Install the Node Feature Discovery Operator from OperatorHub in the OKD console.

  2. After installing the NFD Operator into OperatorHub, select Node Feature Discovery from the installed Operators list and select Create instance. This installs the nfd-master and nfd-worker pods, one nfd-worker pod for each compute node, in the openshift-nfd namespace.

  3. Verify that the Operator is installed and running by running the following command:

    $ oc get pods -n openshift-nfd
    Example output
    NAME                                       READY    STATUS     RESTARTS   AGE
    
    nfd-controller-manager-8646fcbb65-x5qgk    2/2      Running 7  (8h ago)   1d
  4. Browse to the installed Oerator in the console and select Create Node Feature Discovery.

  5. Select Create to build a NFD custom resource. This creates NFD pods in the openshift-nfd namespace that poll the OKD nodes for hardware resources and catalogue them.

Verification
  1. After a successful build, verify that a NFD pod is running on each nodes by running the following command:

    $ oc get pods -n openshift-nfd
    Example output
    NAME                                       READY   STATUS      RESTARTS        AGE
    nfd-controller-manager-8646fcbb65-x5qgk    2/2     Running     7 (8h ago)      12d
    nfd-master-769656c4cb-w9vrv                1/1     Running     0               12d
    nfd-worker-qjxb2                           1/1     Running     3 (3d14h ago)   12d
    nfd-worker-xtz9b                           1/1     Running     5 (3d14h ago)   12d

    The NFD Operator uses vendor PCI IDs to identify hardware in a node. NVIDIA uses the PCI ID 10de.

  2. View the NVIDIA GPU discovered by the NFD Operator by running the following command:

    $ oc describe node ip-10-0-132-138.us-east-2.compute.internal | egrep 'Roles|pci'
    Example output
    Roles: worker
    
    feature.node.kubernetes.io/pci-1013.present=true
    
    feature.node.kubernetes.io/pci-10de.present=true
    
    feature.node.kubernetes.io/pci-1d0f.present=true

    10de appears in the node feature list for the GPU-enabled node. This mean the NFD Operator correctly identified the node from the GPU-enabled MachineSet.

Enabling Accelerated Networking on an existing Microsoft Azure cluster

You can enable Accelerated Networking on Azure by adding acceleratedNetworking to your machine set YAML file.

Prerequisites
  • Have an existing Microsoft Azure cluster where the Machine API is operational.

Procedure
  • Add the following to the providerSpec field:

    providerSpec:
      value:
        acceleratedNetworking: true (1)
        vmSize: <azure-vm-size> (2)
    1 This line enables Accelerated Networking.
    2 Specify an Azure VM size that includes at least four vCPUs. For information about VM sizes, see Microsoft Azure documentation.
Next steps
  • To enable the feature on currently running nodes, you must replace each existing machine. This can be done for each machine individually, or by scaling the replicas down to zero, and then scaling back up to your desired number of replicas.

Verification
  • On the Microsoft Azure portal, review the Networking settings page for a machine provisioned by the machine set, and verify that the Accelerated networking field is set to Enabled.