VirtIO dropping packets but e1000e works fine

moored99

Member
Jul 27, 2023
10
0
6
I was warned by cPanel support that my cPanle VM was dropping packets.

I did some testing and setup the VM with the Intel e1000e network which did not show any errors

Proxmox 9.2.9
VM setup as SeaBIOS
Machine i440fx
SCSI SCSI VirtIO SCSI Single

Using VirtIO driver
ens18: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 9000
inet 10.14.35.1 netmask 255.255.255.0 broadcast 10.14.35.255
inet6 fe80::be24:11ff:feb1:c174 prefixlen 64 scopeid 0x20<link>
ether bc:24:11:b1:c1:74 txqueuelen 1000 (Ethernet)
RX packets 70401711 bytes 42709561655 (39.7 GiB)
RX errors 0 dropped 328280 overruns 0 frame 0
TX packets 78937181 bytes 181835223297 (169.3 GiB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0

Using Intel e1000e driver
ens18: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 9000
inet 10.14.35.1 netmask 255.255.255.0 broadcast 10.14.35.255
inet6 fe80::be24:11ff:feb1:c174 prefixlen 64 scopeid 0x20<link>
ether bc:24:11:b1:c1:74 txqueuelen 1000 (Ethernet)
RX packets 317149 bytes 89008338 (84.8 MiB)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 327713 bytes 610686390 (582.3 MiB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
device interrupt 10 memory 0xfea40000-fea60000

I am using 10GB networking, will using the e1000e have a performance impact ? its Sunday and the server is not under load at the moment.
Is there a fix for the VirtIO driver.
 
How do I find this out?
I apologise, I miss interpreted the log, for Linux vm virtaio driver not an issue (mostly).

I would consider using Virtio and investigating further, e.g:
Check journalctl log in pve and vm.
consider enabling multiqueue.
Make sure the path supports 9000 mtu.

Remember, RX dropped does not necesarily mean a NIC issue. It shows packets were received but dropped before being fully processed.

TX dropped means packets queued for transmission were dropped before being sent.

So this alone does not indicate a hardware/NIC problem.
 
Last edited: