High snapshot stun times

Products

• VMware vSphere ESXi

Issue/Introduction

This document talks about Snapshot Operation limitations directly correlated to number of VMDKs  

Verification Steps 

• Test VM that was used for test is "EXAMPLEVM"
  
  
• "EXAMPLEVM" VM has 4 disks attached to it

scsi0:0 = "EXAMPLEVM.vmdk"
scsi0:1 = "EXAMPLEVM_1.vmdk"
scsi0:2 = "EXAMPLEVM_2.vmdk"
scsi0:3 = "EXAMPLEVM_3.vmdk"

• Normal snapshot creation task in vmware.log of the VM

 Create Snapshot: Test, memory=false, quiescent=false state=5 

• Total Stun time was 245ms 

Checkpoint_Unstun: vm stopped for 245344 us 

• Network ping drops observed. 
  
  
• Some Linux clustering VMs which are IO sensitive may experience frequent cluster failover or at times reboots the VM.

Environment

VMware vSphere ESXi 8.x
VMware vSphere ESXi 7.x
VMware vSphere ESXi 6.x

 

Cause

• To create a VM snapshot, the VM undergoes a temporary “stun” process to:
  • Serialize the device state to disk
  • Close the active disk and establish a snapshot point
• This procedure remains consistent across both Windows and Linux environments.
• The snapshot workflow follows a linear structure, where the duration of the VM stun operation for snapshot creation is directly correlated with the number of VMDKs associated with the virtual machine. 

Resolution

VM snapshot stun times correlate with the number of virtual disks

The above KB explains the concept in detail and point it to the best practices.