High workload Windows VMs report PVSCSI-alert-(129)

Products

• VMware vSphere ESXi

Issue/Introduction

High-intensity database Virtual Machines (VMs) experience intermittent I/O timeouts during heavy processing bursts.

• Guest OS (Windows): Event Viewer reports Source: pvscsi, Event ID: 129 ("Reset to device, \Device\RaidPortX, was issued") occurring at ~60-second intervals.
  

• ESXi Host: vmkernel.log confirms resets with entries: VSCSI: processing reset for handle and IO was aborted by VMFS via a virt-reset.

• Storage Layer: Physical fabric and VMFS-6 datastores show no latency, path failures (APD/PDL), or hardware errors via esxcli storage core device stats.

Environment

ESXi (All versions)
Windows guest OS

Cause

The issue is caused by Virtual SCSI (PVSCSI) queue exhaustion.

When a VM generates a massive I/O burst, the default PVSCSI ring pages (8) and Guest driver queue depth are insufficient to buffer the requests. Once the queue is saturated, the Windows driver exceeds its 60-second I/O timeout and triggers a virt-reset.

The virt-reset log entry indicates the command originated from the Guest OS driver, not the storage array. This confirms the bottleneck is internal to the PVSCSI layer rather than the physical SAN.

Resolution

1. Increase PVSCSI Ring Pages (Guest OS)

Increasing the RequestRingPages allows the driver to handle larger bursts before hitting a timeout.

• • Action: Modify the Windows Registry to increase MaxQueueDepth and RingPages.

  • Registry Path: HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\pvscsi\Parameters\Device

  • Value: MaxQueueDepth=254;RingPages=32

  • Reference: See Broadcom Article 343323 for detailed value recommendations based on workload.

2. Adjust HBA Queue Depth (ESXi Host)

If I/O pressure moves to the physical HBA, increase the module queue depth to ensure the host can process the increased volume.

• • Action: Use esxcfg-module to set the queue depth for the physical adapter (e.g., Emulex lpfc or QLogic qla2xxx).

  • Reference: Refer to the Host-side section of Broadcom Article 343323.

3. Consultation

Microsoft Support: Rule out OS-specific filesystem issues or known cosmetic reporting bugs (e.g., Storport.sys issues).

Additional Information

• Broadcom Article 315143: Instructions for initial PVSCSI adapter configuration and OS compatibility.
• Reversibility: All Guest Registry and Host esxcfg-module changes are fully reversible if unforeseen issues occur.