AZ2 has 4 VMKs and AZ1 has 2 VMKs for nsx-overlay DVS after stretched cluster deployment

Products

• VMware Cloud Foundation

Issue/Introduction

Symptoms:

• In VCF 4.3 and above environments after creating a stretched cluster using JSON, AZ2 hosts have 4 TEP VMKs compared to AZ1 just having 2.
• Comparing to older stretched cluster, equal number of VMKs supposed to exist both on AZ1 and AZ2 hosts.
• No errors observed during creation workflow.
• 3 VMKs one for MGMT, vMotion and VSAN will get created as expected, But the workflow creates other portgroups in AZ2 with another uplink profile with 4 uplinks and 4 overlay VMKs.
• AZ1 uplink would be created during the initial VI/Cluster creation. Only the AZ2 uplink gets created during the stretch workflow 

Environment

VMware Cloud Foundation 4.4.x
VMware Cloud Foundation 4.3

Cause

This issue happens only with multi-pNIC / non-standard deployments, the code picks up the number of VMNIC mappings added per host as per the specified JSON file and creates that many uplinks in the AZ2 uplink profile for the overlay.

Resolution

This is a known issue with VCF 4.3 and is fixed as part of 4.5

Workaround:
The extra uplinks created due to the bug can be removed manually from the NSX-T manager's default teaming policy of AZ2, there is no impact. 

Note: There might be a difference in VMK numbering on AZ1 and AZ2 after the removal of uplinks on NSX this inconsistency will not cause any impact on the VCF workflows

Additional Information

• The VCF stretch workflow doesn't touch/update the uplink profile that was created during the initial cluster/VI creation
• Default uplink profiles present there are the ones created out of the box after NSX installation, they are not used by VCF. VCF creates new uplink profiles for its use. There is no need to edit the default uplink profiles found with 4 uplinks as they are not relevant.

Impact/Risks:
None