Search results

You're right. Any idea why it would bog down the OS with host/nested running? I don't mean like less performance. It takes a minute to load windows and any game runs at ~15 FPS compared to 140+ with the kvm cpu selected. I have another install running without proxmox and do not have that issue...

Are you sure? I'm running windows with hyper-v enabled. When hyper-v is enabled the host os also runs on top of the Hyper-V virtualization layer, just as guest operating systems do. balloon: 0 bios: ovmf boot: order=sata0 cores: 4 efidisk0: Medium1:vm-101-disk-1,size=4M hostpci0...

When using host as the CPU type nested virtualization (hyper-v) is slow. But when the CPU is set to kvm64 everything is smooth as butter. Anyone have any idea? This is with the 3900x.

I passed through a USB controller and it works now.

Yeah it's not being detected by the windows VM at all. Unplugging and plugging back in does nothing. My cheap usb microphone works though so it's some kind of problem with proxmox.

I'm using vfio and running windows 10 on proxmox. My blue yeti shows up when assigning the usb port but when windows 10 is running proxmox says its disconnected but the microphone has power. I'm not sure what you need to debug this so toss me some requests and I'll get the data for you.

I had to reinstall and once the reinstall finished none of my existing thin pools are being added to storage. It's also not letting me manually add them. https://i.imgur.com/54UePee.png https://i.imgur.com/HoiYFmW.png https://i.imgur.com/kfksEVN.png How can I fix this?

If you're asking about virtio vs QXL though virtio uses less CPU and performs just as good as QXL from my experience. Edit: Especially when you remove the 1Gbps network bottleneck.

Novnc is good but SPICE is higher performing with better framerates IMO. I'm just curious on why the option isn't there.

If I select SPICE it defaults to the QXL driver. Using virt-manager I can select SPICE and use virtio as the display driver.

When using the spice display it requires QXL. It will not work with the virtio driver selected. See here: https://i.imgur.com/5L6Bb3F.png https://i.imgur.com/pQ6h3EP.png

Please allow the virtio driver to use the spice display server. I know this is possible because it works with virt-manager. It does work over novnc even at high 4k resolutions but I prefer the spice experience better.

Yes, it was. But it looks like this may have been an issue with my bios. I upgraded the bios and everything works now.

I'm not sure if VE 6.3 can do it. I've been trying as well with no success but in my case the drivers do install and device manager shows the GPU with everything okay. I don't really want to swap this node back to virt-manager but this breaks my use case for proxmox on this specific node.

My 2000 series nvidia GPU shows up correctly in windows 10's device manager and the nvidia driver installs. However, I can't get any video output from the card no matter what I do. Have there been changes with the VE 6.3 update? To be fair this is a 3900x, but if the card is showing up in...

I am using a separate network. The issue is any changes I made to corosync reverted back on upgrade. And since I'm replacing the 2x10gbps cards with IB it's going to become unusable if it reverts the changes again.

Hi all, I've manually modified corosync to new IP addresses before but they reverted during the last upgrade. I'm going to be changing from 1x1gbps network w/2x10gbps nic to 1x2gbps LACP w/1x40Gbps IB per node. I want the nodes communicating over the 2 Gbps LACP and not IB. Is there a way to...

Nope, tried that.

This was actually a pretty weird issue that required multiple node restarts after moving the OSD. Both the old node and new node had to be down, and IIRC if the old node was still running when restarting the new node it still wouldn't detect the osd correctly. So likely an issue with quarium or...

Using Linux virtio yields poor performance (~30MB/s capped) via lan/wan. Using latest proxmox. It does not matter if it's gentoo or debian - 30MB/s max. sysctl changes appear to have no affect, including disabling offloading on guest/host. INCLUDING host -> guest LAN.