Storage Issues

• vSphere vMotion operations of virtual machines residing on Pure-backed vSphere Virtual Volumes storage might time out

  vSphere vMotion operations for VMs residing on vSphere Virtual Volumes datastores depend on the vSphere API for Storage Awareness (VASA) provider and the timing of VASA operations to complete. In rare cases, and under specific conditions when the VASA provider is under heavy load, response time from a Pure VASA provider might cause ESXi to exceed the timeout limit of 120 sec for each phase of vSphere vMotion tasks. In environments with multiple stretched storage containers you might see further delays in the Pure VASA provider response. As a result, running vSphere vMotion tasks time out and cannot complete.

  Workaround: Reduce parallel workflows, especially on Pure storage on vSphere Virtual Volumes datastores exposed from the same VASA provider, and retry the vSphere vMotion task.

• In a vSphere Virtual Volumes stretched storage cluster environment, some VMs might fail to power on after recovering from a cluster-wide APD

  In high scale Virtual Volumes stretched storage cluster environments, after recovering from a cluster-wide APD, due to the high load during the recovery some VMs might fail to power on even though the datastores and protocol endpoints are online and accessible.

  Workaround: Migrate the affected VMs to a different ESXi host and power on the VMs.

• You see „Object or item referred not found” error for tasks on a First Class Disk (FCD)

  Due to a rare storage issue, during the creation of a snapshot of an attached FCD, the disk might be deleted from the Managed Virtual Disk Catalog. If you do not reconcile the Managed Virtual Disk Catalog, all consecutive operations on such a FCD fail with the Object or item referred not found error.

  Workaround: See Reconciling Discrepancies in the Managed Virtual Disk Catalog.

Installation, Upgrade, and Migration Issues

• Fresh installation or creating VMFS partitions on a Micron 7500 or Intel D5-P5336 NVMe drives might fail with a purple diagnostic screen

  UNMAP commands enable ESXi hosts to release storage space that is mapped to data deleted from the host. In NVMe, the equivalent of UNMAP commands is a deallocate DSM request. Micron 7500 and Intel D5-P5336 devices advertise a very large value in one of the deallocate limit attributes, DMSRL, which is the maximum number of logical blocks in a single range for a Dataset Management command. This leads to an integer overflow when the ESXi unmap split code converts number of blocks to number of bytes, which in turn might cause a failure of either installation or VMFS creation. You see a purple diagnostics screen with an error such as Exception 14 or corruption in dlmalloc. The issue affects ESXi 8.0 Update 2 and later.

  Workaround: For a fresh ESXi installation, first install ESXi 8.0, deactivate UNMAP for the affected disk by using the command esxcli storage core device vaai status set -D 0 -d <device-id> and then upgrade to 8.0 Update 3. To create VMFS partitions, disable UNMAP for the affected disk by using the command esxcli storage core device vaai status set -D 0 -d <device-id> and then create VMFS as usual. You can reactivate UNMAP after you create a VMFS datastore, but if you delete or create a partition, UNMAP must remain deactivated.

• The Cancel option in an interactive ESXi installation might not work as expected

  Due to an update of the Python library, the Cancel option by pressing the ESC button in an interactive ESXi installation might not work as expected. The issue occurs only in interactive installations, not in scripted or upgrade scenarios.

  Workaround: Press the ESC key twice and then press any other key to activate the Cancel option.

• Standard image profiles for ESXi 8.0 Update 3 show last modified date as release date

  The Release Date field of the standard image profile for ESXi 8.0 Update 3 shows the Last Modified Date value. The issue is only applicable to the image profiles used in Auto Deploy or ESXCLI. Base images used in vSphere Lifecyce Manager workflows display the release date correctly. This issue has no functional impact. The side effect is that if you search for profiles by release date, the profile does not show with the actual release date.

  Workaround: Search by release version, such as 8.0.3.