VCF Automation may not function properly when running on ESX 8.0 / 8.0u1 / 8.0u2

Products

• VCF Automation

Issue/Introduction

When VCF Automation is deployed on ESX hosts running version 8.0, 8.0u1, or 8.0u2, the RabbitMQ component may restart unexpectedly, potentially preventing the creation and management of workloads in VCF Automation

Environment

• VCF Automation 9.0
• ESXi 8.0, 8.0u1, or 8.0u2

Cause

There is a known issue in ESXi 8.0 / ESXi 8.0u1 / ESXi 8.0u2 with vSAN 8 that can cause data inconsistencies and checksum failures in applications that use AVX-512 CPU instructions, which impacts the RabbitMQ component of VCF Automation.

 

Resolution

The issue is fixed in ESXi 8.0U2c and later

If upgrade is not possible, following these steps to disable AVX-512 CPU instructions on VCF Automation appliances:

1. In the vSphere console, navigate to "VMs and Templates".
2. Expand the Datacenter and Folder that were configured when VCF-A was deployed and select the template with name that starts with "vcf-services-platform-template".
3. Select "Actions -> Convert to Virtual Machine".
4. Once the conversion completes, select "Actions -> Edit Settings".
5. On the Advanced Parameters tab, add the following two values:
   1. Attribute = featMask.vm.cpuid.AVX512F, Value = Max:0
   2. Attribute = featMask.vm.cpuid.AVX512FP16, Value = Max:0
6. Click OK.
7. Select "Actions -> Template -> Convert to Template".
8. In VCF Operations, Power Off the VCF Automation Deployment.
9. Once Power Off is complete, Power On the VCF Automation Deployment.

Additional Information

Refer to https://knowledge.broadcom.com/external/article/367589 on how to disable AVX-512 CPU instruction in Db2 application