What is the effect of a non-graceful virtual machine shutdown on a virtualized server that is connected to a SAN with Fibre Channel interfaces?

In a virtualized Fibre Channel SAN environment using N-Port ID Virtualization (NPIV), each virtual machine can have its own virtual World Wide Port Name (WWPN). A non-graceful VM shutdown (e.g., a crash) prevents the VM from sending a Fabric Logout (LOGO) command to the SAN switch. This results in a stale login session for the VM's WWPN remaining in the switch's name server database. This stale entry can prevent the VM from successfully logging back into the fabric after a restart and, in some cases, may trigger port security features on the switch that place the associated virtual or physical interface into an error-disabled or non-operational state, rendering it "out-of-service" to maintain fabric stability.

A. A non-graceful VM shutdown is a logical event within the hypervisor; it does not cause the physical server HBA (NP Port) to reload.
C. While in-flight I/O frames for the VM may be orphaned at the storage system, the storage array will eventually time out these requests. This is a secondary effect, not a direct impact on the fabric interface state.
D. Frames in the server's HBA buffer are managed and cleared by the hypervisor and HBA driver. This is an internal server process, not a direct effect on the external SAN switch.

1. Cisco MDS 9000 Series NX-OS Troubleshooting Guide, Release 8.x, "Troubleshooting N-Port Virtualization" section. This guide discusses issues arising from stale FLOGI entries when devices do not log out gracefully. It notes that a stale entry can prevent a device from logging in again until the entry is cleared, effectively making the connection for that WWPN non-operational.

2. Cisco, "Troubleshooting N-Port ID Virtualization," Document ID: 116483, (Updated Jan 28, 2014). This document explicitly states, "A common NPIV problem is a stale entry in the FLOGI database on the switch. This occurs when a vHBA is removed ungracefully and the switch does not receive a LOGO. The switch holds the FLOGI database entry for that vHBA until the fctimer values are met." This stale entry is the root cause of the interface becoming unusable.

3. VMware vSphere Documentation, "vSphere Storage," for ESXi 7.0, Chapter 11: "Fibre Channel and vSphere," Section: "N-Port ID Virtualization." The documentation explains that during a graceful shutdown, the hypervisor sends a LOGO to the fabric. It implicitly confirms that in a non-graceful shutdown, this does not occur, leading to potential login issues that require the switch to time out the old session.