Some Storage Arrays does not support Automatic Guest Unmap using VVOLs (virtual volumes)

Products

• VMware vSphere ESXi

Issue/Introduction

Symptoms:

• Attempting to use thin provisioned VVOLs will not allow space reclamation.
• With thin VVols, UNMAPs from guest do not get passed through to the array when a thin VVol is in use.
• In the vmkernel.log we see similar to below 

          YYYY-MM-DDTHH:MM:SS.152Z cpu30:1001396969)DevFS: Ioctl:5597: Could not get unmap granularity from device 'rfc4122.90769fc6-1a5f-4e48-a6ef-f25b4898d49f': Not supported
          YYYY-MM-DDTHH:MM:SS.158Z cpu20:1001396969)DevFS: Ioctl:5597: Could not get unmap granularity from device 'rfc4122.90769fc6-1a5f-4e48-a6ef-f25b4898d49f': Not supported

Environment

VMware vSphere ESXi 6.7
VMware vSphere ESXi 6.0
VMware vSphere ESXi 6.5

Cause

When a VVOL device is bound its associated the SCSI PE device (all secondary level luns are mapped to the PE that they are bound to). The PE device is the one on which the INQUIRY for VPD 0x50 and 0x52 are done to report if the device supports unmap and the unmap block limits. The block limits are mapped to the PE device as a cache of the values returned for the SCSI device (PE). With some arrays it isn't supporting this VPD for the PE device.

Resolution

This is not a VMware issue. For more information please contact the Storage vendor.

Workaround:

• Use VMFS6 (filesystem based storage)  to gain UNMAP capability.