NSX installation is skipped on stateless host after reboot

Products

• VMware NSX

Issue/Introduction

• In stateless host environments using vSphere Auto Deploy, the NSX VIB is not installed automatically when a host reboots.
  
  
• The host preparation (hostprep) process is ignored during reboot, preventing NSX from installing required components on the host.
• This causes the host to remain unprepared for NSX until manual intervention occurs, such as pressing the RESOLVE button in the NSX UI.
  
  
• On the NSX Manager logs /var/log/proton/nsxapi.log , we see the following log entry:

  YYYY-MM-DDTHH:MM:SS.NNN  INFO DiscoveredNodeStreamListener-1-1 TransportNodeCollectionServiceImpl ##### FABRIC [nsx@#### comp="nsx-manager" level="INFO" subcomp="manager"] Ignore hostprep dcn as TNP is already applied on host: <Host_UUID> by cluster: <ComputeCollectionId>
  YYYY-MM-DDTHH:MM:SS.NNN  INFO DiscoveredNodeStreamListener-1-1 DiscoveredNodeStreamListener ##### FABRIC [nsx@#### comp="nsx-manager" level="INFO" subcomp="manager"] DCN Handling for Stateless ESX power-on event : skipping event for DiscoveredNodeProperties[<ComputeCollection_UUID>:<Host_MOID>] with values [OperationType:UPDATE, isEligibleForHostprep=false]

        As per above log "Ignore hostprep dcn as TNP is already applied on host:<Host_UUID> by cluster: <ComputeCollectionId>" and "isEligibleForHostprep=false" this is seen when the host prep is going on through apply-TNP-on-the-cluster workflow. 

 

Environment

NSX 4.2.2.X

Cause

Stale entries remain in the Transport Node Collection(TNC) installation table which causes the product to incorrectly determine installation is in progress on the cluster when it is not.

Resolution

This issue is fixed in NSX 4.2.3.
In  NSX 4.2.2, the workaround where installation is completely missed is to manually remove the stale entry from the database.
Please contact the Broadcom support team for instructions on how to remove this.