Virtual machine disks become corrupted after storage vMotion between two different arrays

Products

• VMware vSphere ESXi

Issue/Introduction

Migrating VMs from one datastore, on one array, to another datastore, on another array. After the Storage vMotion, the VMs do not power on and boot disk is corrupted.

Migrating the same virtual machine to another datastore on the same array does not corrupt the disks.   

Hardware acceleration (VAAI) is enabled in the cluster.

Environment

vSphere 7.0.X

vSphere 8.0.X

Cause

The the storage systems are configured to comply with the request to support storage array based migration.

This requires it to take over the virtual ID of the storage system. If there is a matching ID on the new array, the XCOPY is seen as completed by the vCenter. 

When VAAI is enabled, some of the verification processes are handed off to the storage array. When this happens, the vCenter does not engage in the proper identification of the block locations on the array for distinguishing if a hardware accelerated XCOPY is possible or not. The blocks are copied over incorrectly and the files become corrupted.

Resolution

In order to resolve this issue, disable hardware acceleration on each host in the cluster. Run the storage vMotion. Then, re-enable hardware acceleration on all of the hosts.

https://knowledge.broadcom.com/external/article/318224/disabling-the-vaai-functionality-in-esxi.html