Performance issues on VMs caused by storage array UNMAP command limits.

Products

• VMware vSphere ESXi

Issue/Introduction

• VMs experience slowness or become completely unresponsive to the users.
  
  
• Operations such as vMotion, snapshot creation, or VM power-on tasks experience abnormal delays or fail with timeout errors.
  
  
• Latency is observed at the application layer, with high disk I/O wait times reported within the Guest OS.

Environment

VMware ESXi 7.x
VMware ESXi 8.x

Cause

The slowness issue in the VMs are caused by storage array reaching its command limit and rejecting unmap IO commands from the ESXi host with a "device busy" error.

SCSI command 0x42 is an unmap command.

SCSI code D:0x8 indicates "device busy".

Reference:
From ESXi, var/run/log/vmkernel.log reports sense code "D:0x8" events.


Device Status  [BUSY] - The status is returned when a LUN cannot accept SCSI commands at the moment. As this should be a temporary condition, the command is tried again.

2025-11-18T23:52:14.824Z cpu11:2097763)NMP: nmp_ThrottleLogForDevice:3867: Cmd 0x42 (0x45d9389dd508, 2486025) to dev "naa.##########################" on path "vmhba64:C0:T0:L2" Failed:
2025-11-18T23:52:14.824Z cpu11:2097763)NMP: nmp_ThrottleLogForDevice:3875: H:0x0 D:0x8 P:0x0 . Act:NONE. cmdId.initiator=0x430688dd5240 CmdSN 0x28bed30
2025-11-22T10:49:15.776Z cpu25:2097777)ScsiDeviceIO: 4115: Cmd(0x45d92f999248) 0x42, CmdSN 0x2d1aa72 from world 2541389 to dev "naa.##########################" failed H:0x0 D:0x8 P:0x0
2025-11-22T10:49:15.777Z cpu25:2097777)NMP: nmp_ThrottleLogForDevice:3798: last error status from device naa.########################## repeated 10 times

Resolution

To mitigate the performance impact caused by excessive UNMAP operations,follow below steps:

• Review the pattern of unmap I/O from ESXi hosts to the array.
• Consider adjusting the level of unmap I/O to avoid overloading the array: Space Reclamation on vSphere VMFS Datastores 

Additional Information

ストレージアレイのUNMAPコマンド制限に起因するVMのパフォーマンス問題