Non-Graceful Virtual Machine Shutdown and its Impact on Fibre Channel SAN

B.

When a virtual machine (VM) is powered off or shutdown in a non-graceful manner, it means that the VM is abruptly stopped without following the proper shutdown procedures, such as closing all running applications, saving data, and shutting down the operating system. In a virtualized environment, this can have various effects on the underlying storage infrastructure, especially if the virtualized server is connected to a Storage Area Network (SAN) using Fibre Channel interfaces.

Fibre Channel is a high-speed storage networking technology that allows servers to access shared storage resources such as disks, tape drives, or solid-state devices. Fibre Channel uses a protocol called SCSI (Small Computer System Interface) to communicate between the server and the storage system. When a VM is running on a virtualized server, it uses Fibre Channel Host Bus Adapters (HBAs) to access the storage resources through the SAN fabric.

If a VM is shutdown non-gracefully, it can cause various issues depending on the state of the Fibre Channel interfaces, the SAN switches, and the storage system. Here are some possible effects:

A. NP Port reload connected to the upstream SAN switch If the Fibre Channel interfaces on the virtualized server are not properly shut down before the VM is powered off, the SAN switch may detect a link failure or a loss of signal on the port connected to the server. This can trigger a Non-Participating (NP) Port reload on the switch, which means that the switch resets the port to a default state and reinitializes the link. This can cause a temporary disruption in the SAN fabric and affect the connectivity of other devices attached to the same switch.

B. Out-of-service Fibre Channel interfaces on the SAN switch If the SAN switch detects a sudden link failure or a loss of signal on a Fibre Channel port, it may mark the port as out-of-service or faulted. This means that the switch disables the port and prevents any traffic from passing through it until the link is restored or manually cleared. If the virtualized server has multiple Fibre Channel interfaces connected to the SAN switch, this can affect only the affected port or all the ports on the server.

C. Frames stuck in the storage system If the VM is running I/O operations (such as reading or writing data) when it is powered off, it may leave some frames (SCSI commands or data blocks) stuck in the storage system. These frames may not be properly completed or acknowledged by the server, and may cause the storage system to wait for a timeout or retry the operation. This can result in degraded performance or errors for other servers or applications that access the same storage resources.

D. Frames stuck in the HBA buffer of the server If the Fibre Channel interfaces on the virtualized server are not properly shut down before the VM is powered off, they may have some frames in their buffers that are not yet transmitted to the SAN fabric. These frames may get stuck in the buffer and prevent other I/O operations from completing. This can also cause the HBA driver to become unresponsive or cause the server to hang or crash.

In summary, a non-graceful VM shutdown can have various effects on a virtualized server connected to a SAN with Fibre Channel interfaces, including triggering a NP Port reload on the upstream SAN switch, marking the Fibre Channel interfaces as out-of-service, leaving frames stuck in the storage system, and causing frames to be stuck in the HBA buffer of the server. To avoid these issues, it is recommended to follow the proper shutdown procedures for the virtualized server and its applications, and to ensure that the Fibre Channel interfaces are properly shut down before the VM is powered off.