ESXi 8.0 U3 hostd hangs and HA fails to trigger on Cisco UCS with nfnic driver 5.0.0.50.

Products

• VMware vSphere ESXi

Issue/Introduction

• After upgrading to ESXi 8.0 U3, hosts running on Cisco UCS hardware may experience management unresponsiveness (hostd hangs) and storage path failures.
• vSphere HA may fail to restart virtual machines because the host remains partially "alive" while the storage and management stacks are deadlocked.
• ESXi host's vmkernel.log will show nfnic_fcpio_itmf_cmpl_handler aborts and Tport discovery failures.

Environment

• Hardware: Cisco UCS B-Series / C-Series

• Software: VMware ESXi 8.0 U3 (Build 24022510 or higher)

• Driver: nfnic 5.0.0.50-1OEM

Cause

• The issue is caused by a mismatch between the nfnic driver and the Cisco VIC firmware.
• Using an older firmware (e.g., 4.x) with the 5.x nfnic driver causes a failure in the FCoE initialization sequence, leading to blocked I/O and kernel-level service hangs.

Resolution

Upgrade the Cisco UCS Firmware to version 5.3(5) or higher as specified in the Broadcom Compatibility Guide. Ensure that both the Infrastructure (A-bundle) and Server (B-bundle) are updated to maintain supportability.

Workaround:

If the ESXi host becomes unresponsive, reboot of the host should recover it. 

Additional Information

Also see KB article "Cisco host experience PSOD event "PF Exception 14" with Module nfnic 5.0.0.4x involved in panic".