Scaling TKG Management Cluster Nodes Vertically

In a previous post I wrote about how to scale workload cluster control plane and worker nodes vertically. This post explains how to do the same for the TKG Management Cluster nodes.

Scaling vertically is increasing or decreasing the CPU, Memory, Disk or changing other things such as the network for the nodes. Using the Cluster API it is possible to make these changes on the fly, Kubernetes will use rolling updates to make the necessary changes.

First change to the TKG Management Cluster context to make the changes.

Scaling Worker Nodes

Run the following to list all the vSphereMachineTemplates.

k get vspheremachinetemplates.infrastructure.cluster.x-k8s.io -A
NAMESPACE    NAME                         AGE
tkg-system   tkg-mgmt-control-plane       20h
tkg-system   tkg-mgmt-worker              20h

These custom resource definitions are immutable so we will need to make a copy of the yaml file and edit it to add a new vSphereMachineTemplate.

k get vspheremachinetemplates.infrastructure.cluster.x-k8s.io -n tkg-system   tkg-mgmt-worker -o yaml > tkg-mgmt-worker-new.yaml

Now edit the new file named tkg-mgmt-worker-new.yaml

apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
kind: VSphereMachineTemplate
metadata:
  annotations:
    kubectl.kubernetes.io/last-applied-configuration: |
      {"apiVersion":"infrastructure.cluster.x-k8s.io/v1beta1","kind":"VSphereMachineTemplate","metadata":{"annotations":{"vmTemplateMoid":"vm-9726"},"name":"tkg-mgmt-worker","namespace":"tkg-system"},"spec":{"template":{"spec":{"cloneMode":"fullClone","datacenter":"/home.local","datastore":"/home.local/datastore/lun01","diskGiB":40,"folder":"/home.local/vm/tkg-vsphere-tkg-mgmt","memoryMiB":8192,"network":{"devices":[{"dhcp4":true,"networkName":"/home.local/network/tkg-mgmt"}]},"numCPUs":2,"resourcePool":"/home.local/host/Management/Resources/tkg-vsphere-tkg-Mgmt","server":"vcenter.vmwire.com","storagePolicyName":"","template":"/home.local/vm/Templates/photon-3-kube-v1.22.9+vmware.1"}}}}
    vmTemplateMoid: vm-9726
  creationTimestamp: "2022-12-23T15:23:56Z"
  generation: 1
  name: tkg-mgmt-worker
  namespace: tkg-system
  ownerReferences:
  - apiVersion: cluster.x-k8s.io/v1beta1
    kind: Cluster
    name: tkg-mgmt
    uid: 9acf6370-64be-40ce-9076-050ab8c6f41f
  resourceVersion: "3069"
  uid: 4a8f305f-0b61-4d33-ba02-7fb3fcc8ba22
spec:
  template:
    spec:
      cloneMode: fullClone
      datacenter: /home.local
      datastore: /home.local/datastore/lun01
      diskGiB: 40
      folder: /home.local/vm/tkg-vsphere-tkg-mgmt
      memoryMiB: 8192
      network:
        devices:
        - dhcp4: true
          networkName: /home.local/network/tkg-mgmt
      numCPUs: 2
      resourcePool: /home.local/host/Management/Resources/tkg-vsphere-tkg-Mgmt
      server: vcenter.vmwire.com
      storagePolicyName: ""
      template: /home.local/vm/Templates/photon-3-kube-v1.22.9+vmware.1

Change the name of the CRD on line 10. Make any other changes you need, such as CPU on line 32 or RAM on line 27. Save the file.

Now you’ll need to create the new vSphereMachineTemplate.

k apply -f tkg-mgmt-worker-new.yaml

Now we’re ready to make the change.

Lets first take a look at the MachineDeployments.

k get machinedeployments.cluster.x-k8s.io -A

NAMESPACE    NAME            CLUSTER    REPLICAS   READY   UPDATED   UNAVAILABLE   PHASE     AGE   VERSION
tkg-system   tkg-mgmt-md-0   tkg-mgmt   2          2       2         0             Running   20h   v1.22.9+vmware.1

Now edit this MachineDeployment.

k edit machinedeployments.cluster.x-k8s.io -n tkg-system   tkg-mgmt-md-0

You need to make the change to the section spec.template.spec.infrastructureRef under line 56.

 53       infrastructureRef:
 54         apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
 55         kind: VSphereMachineTemplate
 56         name: tkg-mgmt-worker

Change line 56 to the new VsphereMachineTemplate CRD we created earlier.

 53       infrastructureRef:
 54         apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
 55         kind: VSphereMachineTemplate
 56         name: tkg-mgmt-worker-new

Save and quit. You’ll notice that a new VM will immediately start being cloned in vCenter. Wait for it to complete, this new VM is the new worker with the updated CPU and memory sizing and it will replace the current worker node. Eventually, after a few minutes, the old worker node will be deleted and you will be left with a new worker node with the updated CPU and RAM specified in the new VSphereMachineTemplate.

Scaling Control Plane Nodes

Scaling the control plane nodes is similar.

k get vspheremachinetemplates.infrastructure.cluster.x-k8s.io -n tkg-system tkg-mgmt-control-plane -o yaml > tkg-mgmt-control-plane-new.yaml

Edit the file and perform the same steps as the worker nodes.

You’ll notice that there is no MachineDeployment for the control plane node for a TKG Management Cluster. Instead we have to edit the CRD named KubeAdmControlPlane.

Run this command

k get kubeadmcontrolplane -A

NAMESPACE    NAME                     CLUSTER    INITIALIZED   API SERVER AVAILABLE   REPLICAS   READY   UPDATED   UNAVAILABLE   AGE   VERSION
tkg-system   tkg-mgmt-control-plane   tkg-mgmt   true          true                   1          1       1         0             21h   v1.22.9+vmware.1

Now we can edit it

k edit kubeadmcontrolplane -n tkg-system   tkg-mgmt-control-plane

Change the section under spec.machineTemplate.infrastructureRef, around line 106.

102   machineTemplate:
103     infrastructureRef:
104       apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
105       kind: VSphereMachineTemplate
106       name: tkg-mgmt-control-plane
107       namespace: tkg-system

Change line 106 to

102   machineTemplate:
103     infrastructureRef:
104       apiVersion: infrastructure.cluster.x-k8s.io/v1beta1
105       kind: VSphereMachineTemplate
106       name: tkg-mgmt-control-plane-new
107       namespace: tkg-system

Save the file. You’ll notice that another VM will start cloning and eventually you’ll have a new control plane node up and running. This new control plane node will replace the older one. It will take longer than the worker node so be patient.

Advertisement

Author: Hugo Phan

@hugophan

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: