VKS worker nodes from the same node pool schedule on same ESXi host

Products

• VMware vSphere Kubernetes Service

Issue/Introduction

In vSphere Kubernetes Service (VKS), worker nodes  from the same node pool within guest clusters may be located on the same physical ESXi host. This behavior bypasses expected high-availability distribution, increasing the risk of simultaneous node failure if a single ESXi host goes down.

Environment

• vSphere Kubernetes Service (VKS)
• vSphere with Tanzu
• vCenter Server 8.0

Cause

VKS does not natively implement or orchestrate vCenter DRS "Separate Virtual Machines" rules for worker nodes.
While DRS manages resource-based placement, it is unaware of the application-level requirement to separate ephemeral Kubernetes nodes unless a rule is manually defined.

Resolution

To ensure ESXi host-level separation, manually configure and maintain DRS Anti-Affinity rules in vCenter.

1. Log in to the vCenter UI.
2. Navigate to the Cluster hosting the VKS cluster.
3. Select the Configure tab and under Configuration, select VM/Host Rules.
4. Click Add.
5. Enter a name for the rule (e.g., vks-nodepool-01-anti-affinity).
6. Set the Type to Separate Virtual Machines.
7. Click Add and select all worker node VMs belonging to the specific node pool.
8. Click OK.

Additional Information

Note on Lifecycle Events:
These rules are not persistent through cluster upgrades or node pool scaling. When a "rolling update" or scaling event occurs, the old VM objects are deleted and replaced. You must manually update the DRS rule to include the new VM objects after these events.