Memory snapshot of the VM when on VVOL is taking more time

Products

• VMware vSphere ESXi

Issue/Introduction

Symptoms:
Memory snapshot of the VM running on a VVOL datastore takes more time compared to when VM is moved to VMFS datastore.

Logs below show that every memory snapshot taken on a VM is quite slow and taking more time

vmware-1.log:2019-08-23T02:16:49.372Z| vmx| I125: SnapshotVMX_TakeSnapshot start: 'VM Snapshot 23%252f08%252f2019, 03:16:48', deviceState=1, lazy=1, quiesced=0, forceNative=0, tryNative=1, saveAllocMaps=0
vmware-1.log:2019-08-23T02:28:18.002Z| vmx| I125: SnapshotVMXTakeSnapshotComplete: Done with snapshot 'VM Snapshot 23%252f08%252f2019, 03:16:48': 1
vmware-1.log:2019-08-23T11:19:26.613Z| vmx| I125: SnapshotVMX_TakeSnapshot start: 'VM Snapshot 23%252f08%252f2019, 12:19:25', deviceState=1, lazy=1, quiesced=0, forceNative=0, tryNative=1, saveAllocMaps=0
vmware-1.log:2019-08-23T11:32:38.908Z| vmx| I125: SnapshotVMXTakeSnapshotComplete: Done with snapshot 'VM Snapshot 23%252f08%252f2019, 12:19:25': 2
vmware-1.log:2019-08-23T20:27:49.878Z| vmx| I125: SnapshotVMX_TakeSnapshot start: 'VM Snapshot 23%252f08%252f2019, 21:27:48', deviceState=1, lazy=1, quiesced=0, forceNative=0, tryNative=1, saveAllocMaps=0
vmware-1.log:2019-08-23T20:41:27.978Z| vmx| I125: SnapshotVMXTakeSnapshotComplete: Done with snapshot 'VM Snapshot 23%252f08%252f2019, 21:27:48': 3
vmware-1.log:2019-08-24T13:56:13.224Z| vmx| I125: SnapshotVMX_TakeSnapshot start: 'VM Snapshot 24%252f08%252f2019, 14:56:10', deviceState=1, lazy=1, quiesced=0, forceNative=0, tryNative=1, saveAllocMaps=0
vmware-1.log:2019-08-24T14:10:49.290Z| vmx| I125: SnapshotVMXTakeSnapshotComplete: Done with snapshot 'VM Snapshot 24%252f08%252f2019, 14:56:10': 4

Environment

VMware vSphere ESXi 7.0.x
VMware vSphere ESXi 6.7

Cause

Low priority I/O throttling is taking place when memory snapshots on VM located on a VVOL datastore

Resolution

An improvement in memory snapshot for SCSI vVols has been implemented and it is included in vSphere 7.0 U3c.



Workaround:
As a workaround you can follow the steps below (make sure to have valid backups before performing any changes):

1. Power off the VM
2. Take a backup of the vmx file.
3. Add the following parameters in vmx configuration file and save.

mainMem.ioBlockPages = "2048"
mainMem.iowait = "1"

4. Reload the vmx file, see:

• Reloading a vmx file without removing the virtual machine from inventory (1026043)

5. Power on the vm.

Additional Information

Impact/Risks:
Memory snapshot of the VM running on a VVOL datastore takes more time compared to when VM is moved to VMFS datastore.