Generating stale WRITE_ZEROES events on Linux virtual machines

Products

• VMware vSphere ESXi

Issue/Introduction

• The following events were observed from the Linux VM.

<vm_name> kernel: blk_update_request: critical target error, dev sda, sector 44147864 op 0x9:(WRITE_ZEROES) flags 0x800 phys_seg 0 prio class 0

• The VMDK is configured using thick provisioning.

Environment

VMware vSphere ESXi 7.x

VMware vSphere ESXi 8.x

Cause

• The events observed from the Linux VM are related to UNMAP/WRITE ZEROES operations on the VMDK disk.

• These operations are intended to reclaim unused space on the disk. In this case, the guest OS attempted to perform a space reclamation, but the process failed due to a misconfiguration.

• To support UNMAP operations, the VMDK must be thin provisioned. Since the disk is currently configured as thick provisioned (Lazy Zeroed), these operations are not supported, which may result in the observed errors.

• Additionally, this behavior can be controlled at the filesystem level by enabling or disabling the discard (i.e., trim/no-trim) option on a per-filesystem or mount point basis.

Resolution

• To use Thick Provisioned VMDKs, you need to disable the trim (discard) option on the relevant filesystems from within the guest OS.

• For guidance on how to do this based on your specific operating system, please engage your OS vendor.