Windows VM NIC disconnects and how to fix it

[alexc]

As we moved to PVE 6.2 (just several hosts, not huge install), we noticed that Windows server VMs we used to run on PVE are to see sudden NIC stop working. The fix is really simple: set NIC to not connected (no link) in PVE web interface, apply changes, the connect it again and apply again - all become fine until next lost of network.

Several VMs with different Windows versions, different Virtio drivers (both E100 and Virtio NICs are tested) - no luck. This shouldn't be that way I suppose.

All hosts are of the same hardware paltform, so host NICs are the same, too, Not tried to replace drivers/modules due to our policy to use pure PVE (no upgrades beside pure Debian system utility upgrade) so hosts should remains the same.

So the first question is how to fix that?

and the second question if I can set NIC to "no link" and then to "link" on PVE side via some API call, so I can do that nightly via cron?

 

[wolfgang]

Hi,

this is not normal behavior. What NICs do you use on these nodes (HW)?
Yes you can set it with an API call

HTTP:	GET /api2/json/nodes/{node}/qemu/{vmid}/config
CLI:	pvesh get /nodes/{node}/qemu/{vmid}/config

 

[alexc]

This is onboard NIC which lists this way:

1c:00.0 Ethernet controller: Intel Corporation Ethernet Connection X722 for 1GbE (rev 09)
1c:00.1 Ethernet controller: Intel Corporation Ethernet Connection X722 for 1GbE (rev 09)

What kind of info should I collect more for you?