Post-upgrade from ESXi 7.0 to 8.0, physical Mellanox ConnectX-5 and ConnectX-6 adapters may experience link instability. Symptoms include persistent link down notifications and switch ports reporting a non-connected state despite successful host-side driver initialization.
This behavior occurs due to physical layer negotiation conflicts between the adapter and the network switch. This article outlines the identification and resolution steps for this condition, specifically addressing Forward Error Correction (FEC) negotiation mismatches.
Symptoms:
vmnic link state following ESXi host reboot.vmkernel logs display repetitive link down notifications for the impacted vmnic.non-connected state.esxcli network nic get output indicates Link Status: Down.nmlx5_core driver and PCIe initialization complete without error.A Forward Error Correction (FEC) negotiation mismatch exists between the Mellanox adapter profile and the physical network switch port configuration.
vmnic.a00148079en_us to verify the compatible FEC settings for the Mellanox adapter firmware version installed on host ####.esxcli network nic get -n vmnic#Link Status reports Up and traffic flows as expected across the uplink.https://support.hpe.com/hpesc/public/docDisplay?docId=a00148079en_us