vCenter Server 8.0 U1 vpxd crashes with error: "Scheme error: '/' failed due to divide-by-zero"

Products

• VMware vCenter Server

Issue/Introduction

• vpxd crashes immediately after starting
• In /var/log/vmware/vpxd/vpxd.log

Scheme error: '/' failed due to divide-by-zero error

Failed to estimate vMotion downtime

• Also, you can see the error below in /var/log/vmware/vpxd/vpxd.log

Scheme error: unbound variable.

Environment

VMware vCenter Server 8.0.1

Cause

This happens when a vMotion is triggered to a vGPU virtual machine (manual or through DRS). The divide-by-zero error occurs when there is no vMotion bandwidth available.

Resolution

This is resolved in vCenter Server 8.0 Update 1b (build number 21860503).

 

Note: Please do not perform the work around 1, if the Host are part of vSAN environment as it's not advisable. 

 

 

Workaround 1

1. Take proper snapshots of the environment.
2. Disconnect all hosts from DB
   
   
   • To disconnect all hosts connected to vCenter:
     
     /opt/vmware/vpostgres/current/bin/psql -d VCDB -U postgres -c "UPDATE vpx_host SET enabled = 0"
     
     
   • Validate the change

/opt/vmware/vpostgres/current/bin/psql -U postgres -d VCDB -c "select id,enabled from vpx_host where enabled <> 0;" 
3. Start the vpxd service
4. Disable DRS once vCenter is accessible
   • Note: Disabling DRS will remove all resource pools within the cluster. Refer to the following KB to retain the resource pools: Retaining resource pools when disabling VMware DRS clusters in the vSphere Web Client
5. Reconnect the hosts from the vCenter UI

 

Workaround 2

1. Power off the VM(s) with a vGPU
2. Restart vCenter services