JordanFromIT

Summary: I am experiencing recurring network connectivity failures on Unraid when using a Realtek RTL8126 5GbE NIC. During the failure, ICMP (ping) continues to work, but TCP-based services such as the Unraid WebUI, SSH, and Docker containers (e.g., Plex) become unreachable. The network link remains up and there are no obvious kernel or system errors, but NIC statistics show increasing RX-side errors and missed packets while the issue is occurring. Attached is a screenshot of the error output, and the diagnostics zip. Environment: Unraid version: 7.2.3 NIC: Realtek RTL8126 (5GbE) root@Mediaserver:~# lspci -nn | grep -i ethernet 0b:00.0 Ethernet controller [0200]: Realtek Semiconductor Co., Ltd. RTL8126 5GbE Controller [10ec:8126] (rev 01) Driver: r8126 Network path: RJ45 5GbE NIC → RJ45-to-SFP+ module → Ubiquiti UDM Pro SFP+ port SFP+ modules tested: Third-party RJ45 SFP+ module Official UniFi RJ45 ↔ SFP+ module (UACC-CM-RJ45-10G) Symptoms: TCP-based access fails (Unraid WebUI, SSH, Docker containers, anything hosted by the server) Temporary / Band-aid Fixes: Disconnect, then reconnect ethernet cable Reboot server Disable UDM Pro port, then re-enable Observed Behavior: While the issue is occurring, NIC RX error and missed packet counters increase continuously. A screenshot of the live ethtool -S output is attached showing the counters incrementing in real time. Interface remains up (no link flaps or renegotiation messages observed) ICMP continues to respond normally CPU and memory usage remain normal Local loopback access (curl to localhost) works during the failure No firewall or conntrack issues observed I have had this server with this motherboard since May of 2025, this issue only started after upgrading from major version 6.x to 7.x a couple months back when 7.x was put into stable release channel. No other major hardware or software changes have occurred other than these 2 events. Troubleshooting Performed: Accessed console via my JetKVM (IP KVM device) to do the following Verified sshd and nginx are listening locally Verified the Unraid WebUI responds via localhost during the failure Checked htop for any obvious hanging or maxing of resources Disabled all NIC offloads (TSO/GSO/GRO/LRO, RX/TX offload) Disabled Ethernet flow control Attempted to increase RX/TX ring buffers (driver appears capped at 1024) Tested multiple RJ45 SFP+ modules, including the official UniFi module for $100 with shipping Toned out cat6 cable - passed UDM Pro port diagnostics show no issues or dropped packets - confirming this is isolated to the NIC/Driver The issue continues to recur despite the above steps. Notes: This failure mode is difficult to diagnose because I cannot recreate it, it has to occur naturally which right now is about once per week. mediaserver-diagnostics-20260122-0039.zip