Vertical node scaling

Applies to: AKS on Azure Stack HCI 22H2, AKS on Windows Server

You can change the size of the virtual machines in a given node pool, to increase the resources available to the node pool in AKS enabled by Azure Arc.

To keep up with app demands in Azure Kubernetes Service (AKS), you might need to adjust the number of nodes that run your workloads. In some cases, scaling a cluster horizontally by adding nodes isn't sufficient to meet the demands from your app for more CPU cores or memory.

Without vertical node scaling, you must redeploy to a new node pool and move the app. This situation might not be ideal in resource limited edge environments. To enable this flexibility, AKS Arc introduces the capability to change the virtual machine (VM) size (SKU) of the VMs in a given node pool.

How vertical node scaling in AKS Arc works

In AKS Arc, target cluster node pools are managed internally as a machineDeployment. One property of a machineDeployment is the VM size (SKU) that was selected when the New-AksHciNodePool command was executed.

To change the node pool to a different VM size (SKU), you can use the Set-AksHciNodePool command for changing the VM size for worker nodes and the Set-AksHciCluster command to change the VM size for control plane nodes.

When you submit the command with the new VM size (SKU), a new machineDeployment for the node pool or cluster is created, replacing the existing machine set. This event triggers an update flow in the underlying deployment system. Similar to an OS or Kubernetes version upgrade, the new machineDeployment uses a rolling update to replace one virtual machine in the node pool or control plane after the other. Each upgrade checks that the old node is correctly cordoned and drained before it's removed.

Example process

The following example illustrates vertical node scaling.

Change the VM size for a Linux worker node pool from 4 cores and 6 GB of memory to 4 cores and 8 GB of memory

First, check the current VM size of the node pool on cluster mycluster. From the output, you can see the VM size is Standard_K8S3_v1:

get-akshcinodepool -clustername mycluster

Status       : {Error, Phase, Details}
ClusterName  : mycluster
NodePoolName : mycluster-linux
Version      : v1.22.4
OsType       : Linux
NodeCount    : 2
VmSize       : Standard_K8S3_v1
Phase        : scaling

Standard_K8S3_v1 in the list of available VM sizes shows that it has 4 cores and 6 GB of memory:

Get-AksHciVmSize

VmSize           CPU MemoryGB
------           --- --------
Default          4   4
Standard_A2_v2   2   4
Standard_A4_v2   4   8
Standard_D2s_v3  2   8
Standard_D4s_v3  4   16
Standard_D8s_v3  8   32
Standard_D16s_v3 16  64
Standard_D32s_v3 32  128
Standard_DS2_v2  2   7
Standard_DS3_v2  2   14
Standard_DS4_v2  8   28
Standard_DS5_v2  16  56
Standard_DS13_v2 8   56
Standard_K8S_v1  4   2
Standard_K8S2_v1 2   2
Standard_K8S3_v1 4   6

The new size you want to set for 4 cores and 8 GB of memory is Standard_A4_v2. To update the node pool mycluster-linux, use the Set-AksHciNodePool cmdlet, which has been updated to accept a -VMsize parameter:

Set-AksHciNodePool -ClusterName mycluster -name mycluster-linux -vmsize Standard_A4_v2

After a few minutes, the process is complete. You can check the result by running Get-AksHciNodePool again and verify that the VmSize is now Standard_A4_v2:

get-akshcinodepool -clustername mycluster

Status       : {Error, Phase, Details}
ClusterName  : mycluster
NodePoolName : mycluster-linux
Version      : v1.22.4
OsType       : Linux
NodeCount    : 2
VmSize       : Standard_A4_v2
Phase        : scaling

Next steps

See the documentation for the updated PowerShell commands:

• Set-AksHciNodePool
• Set-AksHciCluster
• Get-AksHciCluster