KVM direct drive access

[Andrei ZeeGiant]

Hello,

I'm looking to setup a new Proxmox 4.4 node using 2x480GB SSD + 2x2TB SAS, and each pair will be a hardware RAID1 array (LSI MegaRaid). The host system and some small LXC containers will reside on the SSD array, however I would like to give a KVM "direct" access to the 2TB RAID1 (KVM will be a baremetal transfer). Unfortunately the dropdown options I'm seeing when adding a drive do not seem to provide this option, and I'm not sure if the details in the dated article found at https://blog.sviluppoweb.eu/2010/07...n-an-image-to-virtualize-it-in-linux-qemukvm/ still apply/would work in this case. The only options I'm seeing are for qcow/raw/vmdk. Can someone please provide some insight? My main concerns with using images such as qcow/raw/vmdk are related to I/O performance, and corruption+repairs.

P.S: I also just ran across http://blog.imnotacyb.org/disk-passthrough-in-proxmox/ which describes what I'm trying to achieve. Would this be possible via the GUI now, or have any drawbacks with future updates or managing the KVM?

 

[wolfgang]

Hi,

Would this be possible via the GUI now

No.

or have any drawbacks with future updates

No should work also in all coming releases.

or managing the KVM

Yes you will never able to make snapshots and migration also not possible.

If you like performance use always virtio as storage bus.

 

[Andrei ZeeGiant]

Hi Wolfgang,

Thanks for the insight! Would other options such as VNC, or start/stop/stats still be available without issues in the web ui? Am I safe to assume the end result in /etc/pve/qemu-server/<kvmid>.conf would be the following considering the scsi ID isn't used and that's the correct partition uuid?

scsi0: /dev/disk/by-uuid/4b22c1ad-b66c-4d7a-b9a6-a9b0527774fd
scsihw: virtio-scsi-pci