vSAN ESXi Host Freeze due to NMI IPI errors
search cancel

vSAN ESXi Host Freeze due to NMI IPI errors

book

Article ID: 440236

calendar_today

Updated On:

Products

VMware vSAN

Issue/Introduction

  • ESXi host becomes unresponsive or "freezes" repeatedly.

  • VMs cannot be migrated off host.

  • Host will not enter Maintenance Mode.

Environment

VMware vSphere ESXi 8.x

Cause

  • Host becomes unresponsive due to failures in physical CPU cores.

  • vobd.log reports the following error for all CPU# on host:

    vobd[###]  [cpuCorrelator] ###us: [vob.cpu.nmi.ipi2.savebt] NMI IPI: PC 0x###, SP 0x### (Src 0x1, CPU#)

The NMI IPI events indicate that a CPU core has become "stuck" or unresponsive for a brief period, forcing the ESXi kernel to issue an NMI to save debug information.
These were occurring across multiple CPUs (CPU0, 1, 2, etc.).

While the host didn't crash immediately, this is a sign of high contention and it can eventually cause the host to go non-responding.

Resolution

  • Reach out to the hardware vendor to do a comprehensive CPU hardware diagnostic check.

  • A hard reboot of the host may be required.  If so please note that as this host is part of a vSAN cluster some data may be in a reduced/inaccessible state.
    Please reference A Host Is Not Responding in vSAN Cluster