Deploying new windows worker nodes in a Tanzu Kubernetes Grid cluster version 1.28 fails because the OVS installation never completes.

Products

• Tanzu Kubernetes Runtime

Issue/Introduction

• The windows worker nodes on version 1.28 and later fail to install OVS by default as a part of the expected default bootstrap.
  
  
• Antrea on the affected worker node comes up without OVS installation and the installation of OVS is only successful when installed manually.
  
  
• Post this within a minute the node goes into a completely unresponsive state unless the management cluster recreates that node.
  
  
• The new node comes back to the same state as mentioned in point number -1 (Antrea installed without OVS). So, the cluster provisioning gets stuck in this vicious cycle.
  
  

• Within the non-working windows worker node, it won't even create the Directory: C:\openvswitch and C:\k\antrea\install_ovs.log for us to review why the OVS installation is failing in the first place. Below command output inside the windows worker node confirms this behaviour.

  PS C:\> cat C:\k\antrea\install_ovs.log

  cat : Cannot find path 'C:\k\antrea\install_ovs.log' because it does not exist.

Environment

Tanzu Kubernetes Grid 

Cause

The Antrea version shipped with the current TKG version is not compatible with the Kubernetes version.

Resolution

The version of Antrea will have to be downgraded in the custom OS image. Contact Broadcom GSS referencing this kb to assist you with the downgrade steps.