Frequently Encountered Issues - Network Troubleshooting in Linux

1 . Scope & Assumptions

• Applies to: RHEL 8/9 (CentOS Stream, Rocky, Alma), Ubuntu 20.04/22.04 LTS, Debian 12, openSUSE Leap 15, and SUSE Linux Enterprise 15.

• Hardware is x86-64 based CPUs with PCIe or SoC NICs (Intel, Broadcom, Mellanox/NVIDIA, Marvell).

• Uses systemd + NetworkManager by default (RHEL/Fedora, Ubuntu desktop) or ifupdown/netplan/systemd-networkd on server flavours.

• User has sudo access and console/KVM or IPMI in case of network loss.

2 . Troubleshooting Workflow (Rule-Out Method)

Proceed layer-by-layer; stop when the fault is found. Details follow.

3 . Physical & Hardware Checks

1. Verify cabling & optics

   • Confirm correct cable type (Cat6A/SFP+/QSFP) and known-good port on the switch.

   • Observe NIC and switch link/activity LEDs. No light → suspect cable, transceiver, or switch port.

2. Check link parameters

   bash

    

   sudo ethtool enp129s0 # replace with your interface

   • Link detected: yes?

   • Speed / Duplex matches expected (e.g., 10000Mb/s, Full).

   • Speed mismatch → autoneg or switch config problem.

3. Validate NIC presence

   bash

    

   lspci -nnk | grep -i -A3 Ethernet

   • Ensure the device and driver in use are listed.

   • Absent device → seating/BIOS/PCIe slot issue.

4. Inspect kernel logs

   bash

    

   sudo dmesg --follow

   • Look for PCI errors (AER: Corrected error), firmware load failures, or link flaps.

5. Firmware & driver updates

   • RHEL/SUSE: vendor DKMS or kmod-<driver> + fwupdate packages.

   • Ubuntu/Debian: apt install firmware-linux-nonfree or vendor scripts (mlxup, bnxt-vf).

4 . Software Configuration Checks

4.1 Identify interface names

4.2 Verify IP settings

• Missing address?

  • NetworkManager: nmcli con show; nmcli con up <name>

  • Netplan (Ubuntu Server ≥18.04): edit /etc/netplan/*.yaml, sudo netplan apply.

  • ifupdown (Debian): edit /etc/network/interfaces, sudo ifup enp129s0.

  • wicked (SUSE): wicked ifstatus eth0.

4.3 DNS resolution

Check /etc/resolv.conf symlink (systemd-resolved vs static file). Mis-configured DNS often looks like network loss.

4.4 VLAN & MTU

Ensure VLAN tags match switch port; jumbo MTU must match end-to-end.

5 . Host Firewall & SELinux/AppArmor

1. List rules:

   • nftables: sudo nft list ruleset

   • firewalld (RHEL, Fedora, openSUSE): sudo firewall-cmd --list-all

   • ufw (Ubuntu): sudo ufw status verbose

2. Temporarily disable for test (replace after!):

   bash

    

   sudo systemctl stop firewalld sudo ufw disable

3. SELinux (RHEL): getenforce → set to Permissive for quick test (sudo setenforce 0).

6 . Connectivity & Routing Tests

7 . Performance Problems

1. Measure throughput & latency

   bash

    

   iperf3 -s # on server iperf3 -c server # on client

2. Detect congestion/offload issues

   bash

    

   ethtool -K enp129s0 ss -plt # check for LISTEN backlog sar -n DEV 1 5 # utilisation spikes

3. Tune

   • Interrupt coalescing: ethtool -C enp129s0 rx-usecs 3 rx-frames 0.

   • Ring buffers: ethtool -G enp129s0 rx 4096 tx 4096.

   • Enable flow-control or RSS for multi-core.

8 . Hardware Isolation Steps

1. Move NIC to different PCIe slot (eliminates backplane issues).

2. Swap cable and switch port.

3. Boot Live-USB of another distro (proves OS install vs hardware).

4. Loopback plug test (fiber loop or twinax).

5. Replace NIC with known-good unit.

6. If dual-port card: test each port separately.

9 . Distro-Specific Notes

10 . When to Escalate

• Repeated kernel oopses or PCIe AER fatal errors → open hardware RMA.

• Driver bug reported in upstream change-log; try latest kernel from distro backport channel.

• Switch shows errors (FCS, CRC, flapping) on port even after cable/NIC swap → escalate to network team.

11 . Quick Reference Command Cheat-Sheet

12 . Conclusion

Start at the physical layer, work upward, and change one variable at a time. Document every step (interface name, IP, switch port, cable ID) so you can roll back and provide clear information if you must open a vendor ticket. Following this structured approach will isolate 90 % + of network faults on Linux workstations and servers.