Proxmox 6.8.12-9-pve kernel has introduced a problem with e1000e Driver and network connection lost after some hours
- Thread starter leandrosgf
- Start date
Just checked, not sure which one, but it was
ethtool -K eno1 gso off gro off tso off tx off rx off rxvlan off txvlan off
Both solutions (Hegse's or the one I posted) do the same basic thing (there may be a few options that are different, but you can change those). What I am not sure about is if the CLI ethtool version will persist across a reboot. I don't think it will, but not 100% sure on that. The interfaces file version will survive a reboot.
An old bug that popped up again indeed. Only started appearing for me after running package updates on the 4th of April, including the kernel. Before I had uptimes of weeks. I'll report back when the fix from the past does it magic once again...
New Kernel version 6.8.12-10 was released. Preliminary first backup test succeeded and I had no freeze of the PVE.
There is also a difference which may play a role - I moved PBS from a VM to a dedicated HW. So far so good. I'll see in a few days if it's stable and report back.
There is also a difference which may play a role - I moved PBS from a VM to a dedicated HW. So far so good. I'll see in a few days if it's stable and report back.
From the Change log:
proxmox-kernel-6.8 (6.8.12-10) bookworm; urgency=medium
* cherry-pick "bnxt_en: Fix GSO type for HW GRO packets on 5750X chips".
* update source and patches to Ubuntu-6.8.0-60.63
proxmox-kernel-6.8 (6.8.12-10) bookworm; urgency=medium
* cherry-pick "bnxt_en: Fix GSO type for HW GRO packets on 5750X chips".
* update source and patches to Ubuntu-6.8.0-60.63
I have run the backup twice today and so far no network issue. Yet again, my setup changed since the original post.
On the other hand, the GSO/GRO fix for Broadcom might have also fixed things for Intel.
Given that GSO (Generic Segmentation Offload) and GRO (Generic Receive Offload) are Linux networking optimizations.
On the other hand, the GSO/GRO fix for Broadcom might have also fixed things for Intel.
Given that GSO (Generic Segmentation Offload) and GRO (Generic Receive Offload) are Linux networking optimizations.
I had to re-apply the ethtool fixes on Linux 6.8.12-10-pve (2025-04-18T07:39Z) as I got more connection losses. Network adapter I219-LM on Dell Micro.
ok, I'll skip this kernel too. Strange that they approve kernels on similar bugs...no words
I hoped you saved my day. Have the same issue since about a few days when I updated my Lenovo ThinkServer. Going back to 6.8.12-9 now and see if it fixes the at least daily hang ups
Yes. Just recognized as the issue returned. But I don´t have this as a backup. Just .10 and .9 and an older 5.x....Kernel, but will this be compatible?
I can confirm that 6-8-12-10-pve didn't fix the problem. It ran for almost 10 hours (and I was happy until now), and then I lost a connection again. I've returned to 6-8-12-8-pve and will wait for the fix. Does anybody know to whom we can address it?