CPU spikes in vvold service: Excessive NVMe Subsystems on ESXi 8.0.x

Products

• VMware vSphere ESXi

Issue/Introduction

ESXi hosts exhibit high CPU utilization (often exceeding 50%) even when no virtual machines (VMs) are powered on or registered to the host. The vvold (vVol Daemon) process is observed as the primary consumer of CPU resources during these spikes.

Symptoms:

• esxtop shows vvold process consuming high CPU.
• CPU utilization remains high after host reboots.
• Issue triggers specifically when certain NVMe-based datastores are mounted.
• The environment uses NVMe/TCP or NVMe/FC.

Environment

• Product: VMware ESXi 8.0.x
• Hardware: Dell PowerEdge R660 (or similar NVMe-capable servers)
• Storage: NVMe/TCP or NVMe/FC configurations
• Daemon: vvold

Cause

The vvold service experiences high overhead when polling or managing an excessive number of NVMe Subsystems. Internal processing thresholds are exceeded when the subsystem count presented to a single ESXi host is higher than 8.

Resolution

To stabilize CPU utilization, align the NVMe configuration with supported limits:

1. Identify the number of NVMe Subsystems currently presented to the impacted ESXi host.
2. Reduce the total number of NVMe Subsystems to a maximum of 8.
3. Rescan storage on the ESXi host.
4. Restart the vvold service to clear the overhead: /etc/init.d/vvold restart
5. Verify CPU utilization in esxtop to confirm vvold usage has returned to normal levels.

For further configuration alignment, refer to: Recommendation for ESXi NVMe/TCP Configuration (KB 378512).

Additional Information

Users experiencing this issue are encouraged to subscribe to this article for any future updates regarding NVMe management optimizations in ESXi.