vMotion of VMs on vSAN cluster fail with the error "The VM failed to resume on the destination during early power on"

Products

• VMware vSAN

Issue/Introduction

• Compute-only vMotion fails for powered-on VMs that reside in a vSAN datastore.
• The error on vCenter Ui is similar to "A general system error occurred: Launch failure ####-##-##T##:##:##.####### The VM failed to resume on the destination during early power on. Module VMMon power on failed. Cause: Module VMMon power on failed. ####-##-##T##:##:##.####### The VM failed to resume on the destination during early power on."
• In the /var/run/log/vmkernel.log, messages similar to the sample below can be seen:

Wa(180) vmkwarning: cpu50:2946733)WARNING: Swap: 3681: Failed to create swap file '/vmfs/volumes/vsan:524########a19cf-7d############5e/d5####66-5##c-4##2-1##c-0c########64/VM_name.vswp
In(182) vmkernel: cpu55:2101060)VmMemXfer: vm 2101060: 2448: Evicting VM with path:/vmfs/volumes/vsan:524########a19cf-7d############5e/d5####66-5##c-4##2-1##c-0c########64/VM_name.vmx
In(182) vmkernel: cpu55:2101060)VmMemXfer: vm 2101060: 2462: Could not find MemXferFS region for /vmfs/volumes/vsan:524########a19cf-7d############5e/d5####66-5##c-4##2-1##c-0c########64/VM_name.vmx

• Upon checking the Skyline Health (Select the cluster in the vCenter inventory > Monitor > vSAN > Skyline Health), it is seen that there are errors related to the Network and "Cluster Partition".
  • In a vSAN standard cluster, data nodes can be partitioned.
  • In a vSAN stretched cluster, data nodes can be partitioned, or the vSAN Witness can be partitioned (isolated) from the cluster.
• All hosts in the cluster are out of maintenance mode and contributing to the vSAN cluster.

Environment

VMware vSAN (All versions).

Cause

This occurs because the vSAN cluster is in a partitioned state and an insufficient number of Fault Domains are currently present in the environment for component placement.

When the VM to be vMotioned is powered on, a new vswap file has to be created, and if there are not enough fault domains to place the vswap components, the vMotion task will fail. The number of fault domains needed depends on the storage policy of the VM.

Note:

When Fault Domains are configured on the cluster, the number of Fault Domains takes precedence over the number of hosts present.

Resolution

Mitigate the partition issue and the vMotion should be successful.

Please refer to:

• • vSAN Health Service - Network Health - vSAN Cluster Partition
  • vSAN Witness appliance partitioned from stretch cluster