On 5/5/21 08:02, aderum...@odiso.com wrote:
Hi Moula,
local device migration is not related to this remote migration serie,
but maybe some improvement could be done.
I'm think about usb device, where we could have the same device on
multiple hosts. (like a security dongle for example).
I think for usb we should be able to detach/migrate/rettach. (it should
work for stateless device, but not a usb drive for example)
Maybe for local nic pci passthroug too.
But for gpu I'm really not sure it's possible, if you have some data in
the local gpu memory, I don't see how it's possible to get it working.
Do you known an hypervisor with this kind of gpu migration
implementation ?
afaik, this is being worked on in qemu/kvm but needs hardware support,
i.e. you can dump the internal state of a supported pci device,
and migrate it to the target
will probably still take a while until normal (non big datacenter)
customers can use this...
_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel
