Supervisor Cluster Update Stalls Due to DRS Partially Automated VM Override Rules

Products

• VMware vSphere Kubernetes Service

Issue/Introduction

Updates to Supervisor Cluster ESXi hosts may fail to proceed if virtual machines (VMs) are pinned to specific hosts due to DRS configurations.
This article provides steps to resolve stalled lifecycle management operations caused by VM-specific automation settings.

Environment

VMware vSphere Kubernetes Service (vSphere with Tanzu)
Any version supported in the current lifecycle management workflow.

Cause

Individual VMs with DRS override rules set to "Partially Automated" prevent vSphere DRS from automatically vacating hosts.
This blocks the lifecycle manager from placing hosts into maintenance mode, which is a requirement for updating the spherelet agent.

Resolution

1. Identify VMs with "Partially Automated" DRS settings on the affected ESXi hosts.
2. Manually migrate these VMs to hosts already updated to the target version.
3. Verify hosts are evacuated and successfully enter maintenance mode.
4. Allow the lifecycle manager to proceed with the spherelet agent update.
5. Post-update, revert or adjust DRS override rules as required by operational policy.

If the issue persists, ensure logs are collected and contact support for further investigation: Contact Broadcom support

Additional Information

For detailed information on configuring DRS automation levels, refer to the vSphere documentation: Set a Custom Automation Level for a Virtual Machine