Network Port Failure

Summary

These tickets cover systems whose expected Ethernet path never comes up, drops intermittently, or disappears from the OS even though the rest of the machine may still boot. The failures span true onboard-NIC or motherboard faults, cable or transceiver mismatches, firmware or CMOS state problems, and cases that only looked like dead ports because the wrong interface or configuration was being used.

Frequency

• 93 tickets

Common Causes

1. Actual onboard NIC or motherboard hardware failure
   Many tickets end with one or more Ethernet ports showing no link, disappearing from enumeration, or failing repeatedly until RMA or workaround hardware was chosen. Examples: #4224, #5602, #17044, #42095, #42655, and 15 more.
2. Firmware, CMOS, or board state corruption
   Some ports returned only after BIOS, IPMI, firmware, or CMOS-reset work, suggesting the network hardware was present but stuck in a bad platform state. Examples: #14586, #15776, #17763, #19140, #21282, and 8 more.
3. Wrong port, management-port confusion, or cabling path mistakes
   A recurring pattern is the customer using the BMC port, the wrong OS-facing jack, or an incompatible cable or transceiver, making the problem look like dead hardware. Examples: #10174, #17152, #18205, #26632, #27247, and 10 more.
4. Driver, OS, or interface-configuration issues
   Several new-system cases were resolved by installing the right driver, fixing netplan or DHCP behavior, or correcting static-IP assumptions rather than replacing hardware. Examples: #13988, #16362, #25210, #40462, #8189, and 12 more.
5. Intermittent PCIe or platform instability presenting as NIC loss
   A smaller but important subset showed NIC errors as part of broader PCIe, DIMM, riser, or post-RMA platform instability. Examples: #18132, #33909, #38928, #38989, #40464, and 7 more.

Diagnostic Steps

1. Confirm you are testing the correct data interface
   Verify the cable is in an OS-facing port, not BMC or management LAN, and confirm the expected adapter is the one being observed. Representative tickets: #10174, #13952, #18205, #26632, #42095.
2. Check whether the NIC is missing, present with NO-CARRIER, or linked but misconfigured
   Use ip a, lspci, lshw -C network, ethtool, and link LEDs to separate enumeration failure from port-link failure and from DHCP or static-IP misconfiguration. Representative tickets: #10916, #17763, #18925, #25210, #27047.
3. Swap simple external variables first
   Change cable, switch port, transceiver, and if relevant try another OS or known-good adapter before assuming the motherboard is dead. Representative tickets: #16362, #17152, #21273, #27247, #40464.
4. Reset platform state when the NIC vanished after reboot or repair
   BIOS updates, CMOS reset, full power drain, and IPMI or BMC updates restored service in several tickets. Representative tickets: #14586, #15776, #17763, #19140, #21282.
5. Escalate to RMA when there is no link in BIOS or the failure recurs after software checks
   Repeated no-link evidence across BIOS and OS usually led to board, barebone, or system repair. Representative tickets: #11736, #17044, #19750, #27390, #41118.

Solutions

1. RMA or replace the failed onboard network path
   The most durable fix in true hardware cases was motherboard, barebone, or full-system repair or replacement. Examples: #11736, #17044, #19750, #27390, #42095, and 14 more.
2. Restore the NIC with CMOS, BIOS, IPMI, or power-drain recovery
   Multiple tickets resolved after resetting stale hardware state rather than replacing the system. Examples: #14586, #15776, #17763, #19140, #21282.
3. Install the correct driver, firmware, or supported transceiver
   This repeatedly solved ports that were physically fine but not usable in the shipped or customer-imaged software stack. Examples: #13988, #16362, #17152, #27401, #40462.
4. Correct cabling or interface selection
   Some of the fastest resolutions came from moving the cable to the correct port or correcting which interface the customer was configuring. Examples: #10174, #18205, #26632, #27247.
5. Use a workaround NIC when downtime blocks RMA
   In a few cases Exxact or the customer kept the machine in service with an add-in NIC while deferring motherboard repair. Examples: #28056, #41686, #42655, #4224.

Edge Cases

• Post-RMA recurrence or dead-on-arrival return: several systems came back from repair still showing NIC or broader platform failures, blurring the line between network-port failure and larger repair-quality issues. See #14081, #18687, #19750, #7270, #8000.
• Intermittent fail-after-hours behavior: some NICs linked at boot and then died after minutes, hours, or days, often pointing to deeper PCIe or board instability rather than a simple dead port. See #24508, #34175, #34258, #38989.
• Networking symptom caused by other failing hardware: one case traced repeated network loss to a bad DIMM, and others tied it to riser or platform faults. See #33909, #40464, #38928.
• Static-IP and bring-up confusion: a subset looked like bad Ethernet hardware but were really static-IP, DHCP, or OS management issues. See #22465, #25210, #30250, #8189.

Related Issues

• nic-hardware-failure
• bios-bmc-issues
• system-boot-failure
• software-installation
• pcie-riser-failure
• motherboard-hardware-failure

Referenced by

• BIOS Firmware Update — co-occurs with this issue (×5)
• Andrew Rodriguez — handled tickets on this issue (×27)
• Garry Gayles — handled tickets on this issue (×8)
• OS Boot Failure — co-occurs with this issue (×4)
• BIOS BMC Issues — co-occurs with this issue (×4)
• System Boot Failure — co-occurs with this issue (×9)
• Duc Bui — handled tickets on this issue (×4)
• H200 — product affected by this issue (×1)
• Nam Luong — handled tickets on this issue (×12)
• H100 — product affected by this issue (×1)