ESXi 7.0.3 Hosts Experience Extreme iSCSI Latency and Disconnects with NetApp Storage

Products

• VMware vSphere ESXi

Issue/Introduction

ESXi 7.0.3 EP13 hosts may experience intermittent storage disconnects and extreme latency (averaging over 1100ms) specifically targeting NetApp iSCSI LUNs. These events lead to datastore instability and degraded host performance in production environments.

• High latency reported on storage access (DAVG/KAVG).
• Intermittent storage path interruptions or "DEAD" path states.
• Temporary datastore unavailability.
• Frequent SCSI sense data entries in vmkernel.log: H:0x0 D:0x2 P:0x0 Valid sense data: 0x6 0x3f 0xe (UNIT ATTENTION - REPORTED LUNS DATA HAS CHANGED)

Environment

• VMware ESXi 7.0.3 (specifically EP13, Build 24585291).
• NetApp iSCSI Shared Storage.
• Intel X710 NICs using the i40en driver.

Cause

The instability is primarily driven by an outdated i40en network driver (v2.2.7.0) interacting with frequent SCSI Unit Attentions triggered by the NetApp storage array.

• Driver Mismatch: The Broadcom Compatibility Guide identifies version 2.5.2.0 as the recommended async driver for Intel X710 on ESXi 7.0 U3. Version 2.2.7.0 lacks critical stability fixes for high-throughput iSCSI traffic.
• SCSI Sense Interpretation: The 0x6 0x3f 0xe Unit Attention indicates LUN inventory or mapping changes on the NetApp side. This forces the ESXi host to perform path discovery. When combined with buffer management inefficiencies in older Intel drivers, this manifests as extreme latency and path cuts.

Resolution

To stabilize the environment, follow these remediation steps:

1. Update Network Drivers: Upgrade the Intel i40en driver to version 2.5.2.0.
2. Apply ESXi Patches: Update hosts to ESXi 7.0.3 EP14 (Build 24784741) or later to benefit from the latest iSCSI stack and pathing improvements.
3. Storage Array Review: Coordinate with the NetApp administrator to identify why "REPORTED LUNS DATA HAS CHANGED" events are being broadcast. Investigate storage-side cluster health, unplanned failovers, or aggregate relocations.

The i40en 2.5.2.0 driver contains critical fixes for frame processing and TSO handling under iSCSI workloads. Patching to EP14 addresses known pathing race conditions in the 7.0.3 branch, and stabilizing array notifications prevents the host from entering frequent, expensive SCSI rescan cycles.