Systemic storage performance degradation, SCSI command aborts, and stall patterns in applications such as backup jobs occur when running RDM (Raw Device Mapping) disks on VMware ESXi hosts with Emulex Fibre Channel adapters. These issues often stem from driver/firmware configuration mismatches between the host adapter and the ESXi hypervisor, leading to severe I/O processing errors.
Symptoms
ScsiDeviceIO performance deterioration warnings appear in vmkwarning logs.An HBA firmware and driver stack mismatch exists. Running uncertified Emulex lpfc stacks in ESXi 8.x environments causes driver worker threads to become overwhelmed during I/O processing (a condition known as "XRI Starvation"), leading to mailbox timeouts and host-wide I/O stuck notifications. RDM configuration exacerbates this by exposing the Guest OS directly to these driver-level aborts, which bypass the typical VMFS buffering layer.
lpfc driver is at version 14.4.576.11 or higher, as this version includes critical fixes for thread deadlock issues.vmkwarning logs post-update to confirm the cessation of latency spikes and command failures.