Bug #416
open
BUG: pfv-tsys1 mgmt interface hangs (e1000/e1000e offload bug)
0%
Description
tsys1 (OptiPlex 9020) management interface intermittently stops responding. Known e1000/e1000e driver bug: hardware unit hang triggered by TSO/GRO/GSO/tx/rx offload on certain Intel NICs under Proxmox.
Fix¶
Disable offload features on the management interface (vmbr0 bridge port NIC):
MAIN_INT=$(brctl show | grep vmbr0 | awk '{print $NF}' | awk -F. '{print $1}')
ethtool -K $MAIN_INT tso off gro off gso off tx off rx off
Persistence¶
Must be applied via systemd service or /etc/network/interfaces post-up to survive reboot.
Existing code¶
Fix script exists at: archive/provisioning/Dell/Server/fixeth.sh
References: https://forum.proxmox.com/threads/e1000-driver-hang.58284/
Priority¶
HIGH — tsys1 hosts critical infra VMs (pfv-bms, pfv-netinfra-01, tsys-ca, tsys-librenms, cnode1).
Scope¶
- Apply offload fix live
- Make persistent across reboots
- Check if other hosts with similar Intel NICs need the same fix
- No reboot required for live fix
Updated by Charles N 2 days ago
Script created at proxmox/perf/scripts/fix-e1000e-offload.sh. Initial deployment attempt to tsys4/5/9 did not execute properly (TSO still on, service not installed). Needs re-deployment with debugging. Awaiting user review/approval before re-attempting.
Affected hosts: tsys4 (nic0), tsys5 (nic1+nic2), tsys9 (nic0). tsys1 and tsys3 already have TSO off.
Updated by Charles N about 5 hours ago
- Due date set to 08/14/2026
- Start date changed from 08/10/2026 to 08/14/2026