e1000e eno1: Detected Hardware Unit Hang:
- Thread starter bogo22
- Start date
I219-V,I218-V chipet drivers are shitty, not only on proxmox, same on xcp-ng for example
https://xcp-ng.org/docs/networking.html#intel-i218-i219-slow-speed
personnaly I'm using mellanox connect-x in production without any problem.
Also wanted to give a big thank you for
This seems to have solved the hardware hangups and DRASTICALLY improved the network latency for the webapps I am running.
I presume that you executed the command in the PVE node, not inside the VM, correct? Moreover, what other guests did you have in the node? I have a mix of VMs and LXCs and this behaviour only happens where I have VMs. Would that be your case, too?
Last edited:
Facing same issues after recent updates on 3 computes and different HW ! seems like something changed after update... system was stable for years before!
see https://bugzilla.proxmox.com/show_bug.cgi?id=6273 - seems like there might have been a regression there
Hi, Fabian. Thanks for the link to the bug tracker. I have contributed to it. I hope it can be solved soon.
I am also having this problem, disable offloading is one way to solve it. Another solution is to modify your VM's network card model to Intel E1000E instead of VirtIO.
I was also able to reliably reproduce the problem with iperf3 client running inside the VM. (The following command might hang your host)
iperf3 -c <IPERF_SERVER> -M 100 -t 50
I was also able to reliably reproduce the problem with iperf3 client running inside the VM. (The following command might hang your host)
iperf3 -c <IPERF_SERVER> -M 100 -t 50
Thanks a lot for your report, @leiwang15 . Much appreciated.
Incidentally, I found this thread on these forums regarding E1000E vs VirtIO on Proxmox.
Fixed this issue on my Proxmox host (currently v8.3.5). Posted here: https://gist.github.com/crypt0rr/60aaabd4a5c29a256b4f276122765237
similar thread (link to my reply which might be useful to folks here): https://forum.proxmox.com/threads/e1000-driver-hang.58284/post-760975
I've changed the VM's NICs to e1000e on 6.8.12-9-pve I219-LM. I'll post the results later.
Also keeping track of this bug report https://bugzilla.proxmox.com/show_bug.cgi?id=6273
Edit:
Bash:
iperf3 -c <IPERF_SERVER> -M 100 -t 50Edit 2:
hang again - using e1000e in the VMs is not fixing the problem!
Last edited:
Yes, turning offloads off for the adapter is the only known to me fix for newer kernels. Currently I apply it manually with:
ethtool -K eno1 gso off tso off rxvlan off txvlan off gro off tx off rx off sg off
Last edited:
This workaround also works for me too, except now I have new messages spammed in dmesg:
Code:
Apr 14 14:01:02 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 14 14:01:27 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.
Apr 14 16:01:08 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 14 16:01:37 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.
Apr 14 18:01:35 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 14 18:01:59 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.
Apr 14 20:00:58 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 14 20:01:25 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.
Apr 14 22:01:15 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 14 22:01:38 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.
Apr 15 00:01:21 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 15 00:01:42 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.
Apr 15 02:01:21 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 15 02:01:46 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.
Apr 15 04:01:23 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 15 04:01:43 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.
Apr 15 06:01:24 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 15 06:01:48 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.
Apr 15 08:01:14 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 15 08:01:39 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.
Apr 15 10:01:23 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 15 10:01:42 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.
Apr 15 12:01:09 pve kernel: EXT4-fs (dm-12): mounted filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5 ro without journal. Quota mode: none.
Apr 15 12:01:34 pve kernel: EXT4-fs (dm-12): unmounting filesystem 987fad4a-ca20-455e-a3c8-c388a14e0cf5.Yeah, I'm actually familiar with that repo—it's my favorite approach, way better than messing with networking directly
I was just hoping this would be fixed by now. Offloading is important!