Intel E1000E 网卡高负载断流

设备是 NUC8beki5 ，装了 PVE 。网卡在网络 IO 高负载时会发生问题，syslog 如下。

现象 1: NUC 网络连接中断，等待几十秒-几分钟后恢复

现象 2: 协商速率由 1000MB/s 变为 100MB/s ，手动设置为 1000MB/s 后恢复 (网卡和路由器都支持千兆)

Jan 27 23:35:35 pve kernel: e1000e 0000:00:1f.6 eno1: NIC Link is Down Jan 27 23:35:35 pve kernel: vmbr0: port 1(eno1) entered disabled state Jan 27 23:35:42 pve kernel: connection1:0: ping timeout of 5 secs expired, recv timeout 5, last rx 10254116869, last ping 10254121984, now 10254127104 Jan 27 23:35:42 pve kernel: connection1:0: detected conn error (1022) Jan 27 23:35:48 pve pvestatd[1233]: status update time (19.582 seconds) Jan 27 23:36:12 pve kernel: e1000e 0000:00:1f.6 eno1: NIC Link is Up 1000 Mbps Full Duplex, Flow Control: Rx/Tx Jan 27 23:36:12 pve kernel: vmbr0: port 1(eno1) entered blocking state Jan 27 23:36:12 pve kernel: vmbr0: port 1(eno1) entered forwarding state Jan 27 23:36:14 pve iscsid[3627392]: connection1:0 is operational after recovery (3 attempts) Jan 27 23:36:37 pve pvestatd[1233]: status update time (19.299 seconds) Jan 27 23:36:38 pve kernel: sd 4:0:0:0: Power-on or device reset occurred 

驱动版本信息:

root@pve:~# ethtool -i eno1 driver: e1000e version: 6.8.12-4-pve firmware-version: 0.4-4 expansion-rom-version: bus-info: 0000:00:1f.6 supports-statistics: yes supports-test: yes supports-eeprom-access: yes supports-register-dump: yes supports-priv-flags: yes 

已经尝试更换网线并按照下面两个帖子的建议关闭了 tso

https://forum.proxmox.com/threads/e1000e-network-issue-on-proxmox.139141/

https://forum.proxmox.com/threads/e1000-driver-hang.58284/page-8#post-390709

但在持续高负载时还会出现上述问题，这个问题是在用 Rsync 恢复备份的时候发现的，尝试用 --bwlimit 限制 Rsync 到 50M, 40M, 30M 。

但均无法解决问题，难道只有硬改网卡了吗？